The following preprints are provided here to allow for a deeper view of our research work, as well as to
promote the rapid dissemination of research results. Please consider, on the other hand, that these preprints
can differ from their published version in ways that may not be entirely negligible, and for this reason we
recommend you to refer to the published version whenever they have to be used or cited.
top2021
Antonucci, A., Facchini, A., Mattei, L. (2021). Structural learning of probabilistic sentential decision diagrams under partial closed-world assumption. In 4th Workshop on Tractable Probabilistic Modeling (TPM 2021 co-located with UAI 2021).
Structural learning of probabilistic sentential decision diagrams under partial closed-world assumption
Authors: Antonucci, A. and Facchini, A. and Mattei, L.
Year: 2021
Abstract: Probabilistic sentential decision diagrams are a class of structured-decomposable probabilistic circuits especially designed to embed logical constraints. To adapt the classical LearnSPN scheme to learn the structure of these models, we propose a new scheme based on a partial closed-world assumption: data implicitly provide the logical base of the circuit. Sum nodes are thus learned by recursively clustering batches in the initial data base, while the partitioning of the variables obeys a given input vtree. Preliminary experiments show that the proposed approach might properly fit training data, and generalize well to test data, provided that these remain consistent with the underlying logical base, that is a relaxation of the training data base.
Published in 4th Workshop on Tractable Probabilistic Modeling (TPM 2021 co-located with UAI 2021).
Structural learning of probabilistic sentential decision diagrams under partial closed-world assumption
@INPROCEEDINGS{antonucci2021e,
title = {Structural learning of probabilistic sentential decision diagrams under partial closed-world assumption},
booktitle = {4th Workshop on Tractable Probabilistic Modeling ({TPM} 2021 {c}o-{l}ocated {w}ith {UAI} 2021)},
author = {Antonucci, A. and Facchini, A. and Mattei, L.},
year = {2021},
doi = {},
url = {https://arxiv.org/abs/2107.12130}
}
DownloadAntonucci, A., Mangili, F., Bonesana, C., Adorni, G. (2021). A new score for adaptive tests in Bayesian and credal networks. In Vejnarová, J., Wilson, N. (Eds), Symbolic and Quantitative Approaches to Reasoning With Uncertainty, Springer International Publishing, Cham, pp. 399–412.
A new score for adaptive tests in Bayesian and credal networks
Authors: Antonucci, A. and Mangili, F. and Bonesana, C. and Adorni, G.
Year: 2021
Abstract: A test is adaptive when the sequence and number of questions is dynamically tuned on the basis of the estimated skills of the taker. Graphical models, such as Bayesian networks, are used for adaptive tests as they allow to model the uncertainty about the questions and the skills in an explainable fashion, especially when coping with multiple skills. A better elicitation of the uncertainty in the question/skills relations can be achieved by interval probabilities. This turns the model into a credal network, thus increasing the inferential complexity of the queries required to select questions. This is especially the case for the information-theoretic quantities used as scores to drive the adaptive mechanism. We present an alternative family of scores, based on the mode of the posterior probabilities, and hence easier to explain. This makes considerably simpler the evaluation in the credal case, without significantly affecting the quality of the adaptive process. Numerical tests on synthetic and real-world data are used to support this claim.
Published in Vejnarová, J., Wilson, N. (Eds), Symbolic and Quantitative Approaches to Reasoning With Uncertainty, Springer International Publishing, Cham, pp. 399–412.
A new score for adaptive tests in Bayesian and credal networks
@INPROCEEDINGS{antonucci2021c,
title = {A new score for adaptive tests in {B}ayesian and credal networks},
editor = {Vejnarov\'a, J. and Wilson, N.},
publisher = {Springer International Publishing},
address = {Cham},
booktitle = {Symbolic and Quantitative Approaches to Reasoning With Uncertainty},
author = {Antonucci, A. and Mangili, F. and Bonesana, C. and Adorni, G.},
pages = {399--412},
year = {2021},
doi = {https://doi.org/10.1007/978-3-030-86772-0_29},
url = {}
}
DownloadAwal, M.R., Cao, R., Lee, R.K., Mitrovic, S. (2021). Angrybert: joint learning target and emotion for hate speech detection. In Karlapalem, K., Cheng, H., Ramakrishnan, N., Agrawal, R.K., Reddy, P.K., Srivastava, J., Chakraborty, T. (Eds), Advances in Knowledge Discovery and Data Mining - 25th Pacific-asia Conference, PAKDD 2021, Virtual Event, May 11-14, 2021, Proceedings, Part I, Lecture Notes in Computer Science 12712, Springer, pp. 701–713.
Angrybert: joint learning target and emotion for hate speech detection
Authors: Awal, M.R. and Cao, R. and Lee, R.K. and Mitrovic, S.
Year: 2021
Published in Karlapalem, K., Cheng, H., Ramakrishnan, N., Agrawal, R.K., Reddy, P.K., Srivastava, J., Chakraborty, T. (Eds), Advances in Knowledge Discovery and Data Mining - 25th Pacific-asia Conference, PAKDD 2021, Virtual Event, May 11-14, 2021, Proceedings, Part I, Lecture Notes in Computer Science 12712, Springer, pp. 701–713.
Angrybert: joint learning target and emotion for hate speech detection
@INPROCEEDINGS{mitrovic2021a,
title = {Angrybert: joint learning target and emotion for hate speech detection},
editor = {Karlapalem, K. and Cheng, H. and Ramakrishnan, N. and Agrawal, R.K. and Reddy, P.K. and Srivastava, J. and Chakraborty, T.},
publisher = {Springer},
series = {Lecture Notes in Computer Science},
volume = {12712},
booktitle = {Advances in Knowledge Discovery and Data Mining - 25th Pacific-{a}sia Conference, {PAKDD} 2021, Virtual Event, May 11-14, 2021, Proceedings, Part I},
author = {Awal, M.R. and Cao, R. and Lee, R.K. and Mitrovic, S.},
pages = {701--713},
year = {2021},
doi = {10.1007/978-3-030-75762-5_55},
url = {}
}
DownloadAzzimonti, D., Rottondi, C., Giusti, A., Tornatore, M., Bianco, A. (2021). Comparison of domain adaptation and active learning techniques for quality of transmission estimation with small-sized training datasets [invited]. IEEE/OSA Journal of Optical Communications and Networking 13(1), pp. A56–A66.
Comparison of domain adaptation and active learning techniques for quality of transmission estimation with small-sized training datasets [invited]
Authors: Azzimonti, D. and Rottondi, C. and Giusti, A. and Tornatore, M. and Bianco, A.
Year: 2021
Abstract: Machine learning (ML) is currently being investigated as an emerging technique to automate quality of transmission (QoT) estimation during lightpath deployment procedures in optical networks. Even though the potential network-resource savings enabled by ML-based QoT estimation has been confirmed in several studies, some practical limitations hinder its adoption in operational network deployments. Among these, the lack of a comprehensive training dataset is recognized as a main limiting factor, especially in the early network deployment phase. In this study, we compare the performance of two ML methodologies explicitly designed to augment small-sized training datasets, namely, active learning (AL) and domain adaptation (DA), for the estimation of the signal-to-noise ratio (SNR) of an unestablished lightpath. This comparison also allows us to provide some guidelines for the adoption of these two techniques at different life stages of a newly deployed optical network infrastructure. Results show that both AL and DA permit us, starting from limited datasets, to reach a QoT estimation capability similar to that achieved by standard supervised learning approaches working on much larger datasets. More specifically, we observe that a few dozen additional samples acquired from selected probe lightpaths already provide significant performance improvement for AL, whereas a few hundred samples gathered from an external network topology are needed in the case of DA.
Published in IEEE/OSA Journal of Optical Communications and Networking 13(1), pp. A56–A66.
Comparison of domain adaptation and active learning techniques for quality of transmission estimation with small-sized training datasets [invited]
@ARTICLE{azzimontid2021,
title = {Comparison of domain adaptation and active learning techniques for quality of transmission estimation with small-sized training datasets [invited]},
journal = {{IEEE/OSA} Journal of Optical Communications and Networking},
volume = {13},
author = {Azzimonti, D. and Rottondi, C. and Giusti, A. and Tornatore, M. and Bianco, A.},
number = {1},
pages = {A56--A66},
year = {2021},
doi = {10.1364/JOCN.401918},
url = {}
}
DownloadBemporad, A., Piga, D. (2021). Global optimization based on active preference learning with radial basis functions. Machine Learning 110, pp. 417–448.
Global optimization based on active preference learning with radial basis functions
Authors: Bemporad, A. and Piga, D.
Year: 2021
Abstract: This paper proposes a method for solving optimization problems in which the decision-maker cannot evaluate the objective function, but rather can only express a preference such as "this is better than that" between two candidate decision vectors. The algorithm described in this paper aims at reaching the global optimizer by iteratively proposing the decision maker a new comparison to make, based on actively learning a surrogate of the latent (unknown and perhaps unquantifiable) objective function from past sampled decision vectors and pairwise preferences. A radial-basis function surrogate is fit via linear or quadratic programming, satisfying if possible the preferences expressed by the decision maker on existing samples. The surrogate is used to propose a new sample of the decision vector for comparison with the current best candidate based on two possible criteria: minimize a combination of the surrogate and an inverse weighting distance function to balance between exploitation of the surrogate and exploration of the decision space, or maximize a function related to the probability that the new candidate will be preferred. Compared to active preference learning based on Bayesian optimization, we show that our approach is competitive in that, within the same number of comparisons, it usually approaches the global optimum more closely and is computationally lighter. Applications of the proposed algorithm to solve a set of benchmark global optimization problems, for multi-objective optimization, and for optimal tuning of a cost-sensitive neural network classifier for object recognition from images are described in the paper. MATLAB and a Python implementations of the algorithms described in the paper are available at http://cse.lab.imtlucca.it/~bemporad/glis.
Published in Machine Learning 110, Springer, pp. 417–448.
Global optimization based on active preference learning with radial basis functions
@ARTICLE{piga2021a,
title = {Global optimization based on active preference learning with radial basis functions},
journal = {Machine Learning},
publisher = {Springer},
volume = {110},
author = {Bemporad, A. and Piga, D.},
pages = {417--448},
year = {2021},
doi = {10.1007/s10994-020-05935-y},
url = {}
}
DownloadBenavoli, A., Azzimonti, D., Piga, D. (2021). Preferential bayesian optimisation with skew gaussian processes. In Proceedings of the Genetic and Evolutionary Computation Conference Companion, GECCO '21, Association for Computing Machinery, New York, NY, USA, 1842–1850.
Preferential bayesian optimisation with skew gaussian processes
Authors: Benavoli, A. and Azzimonti, D. and Piga, D.
Year: 2021
Abstract: Preferential Bayesian optimisation (PBO) deals with optimisation problems where the objective function can only be accessed via preference judgments, such as "this is better than that" between two candidate solutions (like in A/B tests). The state-of-the-art approach to PBO uses a Gaussian process to model the preference function and a Bernoulli likelihood to model the observed pair-wise comparisons. Laplace's method is then employed to compute posterior inferences and, in particular, to build an appropriate acquisition function. In this paper, we prove that the true posterior distribution of the preference function is a Skew Gaussian Process (SkewGP), with highly skewed pairwise marginals and, thus, show that Laplace's method usually provides a very poor approximation. We then derive an efficient method to compute the exact SkewGP posterior and use it as surrogate model for PBO employing standard acquisition functions (Upper-Credible-Bound, etc.). We illustrate the benefits of our exact PBO-SkewGP in a variety of experiments, by showing that it consistently outperforms PBO based on Laplace's approximation both in terms of convergence speed and computational time. We also show that our framework can be extended to deal with mixed preferential-categorical BO, where binary judgments (valid or non-valid) together with preference judgments are available.
Published in Proceedings of the Genetic and Evolutionary Computation Conference Companion, GECCO '21, Association for Computing Machinery, New York, NY, USA, 1842–1850.
Preferential bayesian optimisation with skew gaussian processes
@INPROCEEDINGS{azzimontid2021b,
title = {Preferential bayesian optimisation with skew gaussian processes},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
series = {GECCO '21},
booktitle = {Proceedings of the Genetic and Evolutionary Computation Conference Companion},
author = {Benavoli, A. and Azzimonti, D. and Piga, D.},
pages = {1842--1850},
year = {2021},
doi = {10.1145/3449726.3463128},
url = {}
}
DownloadBenavoli, A., Azzimonti, D., Piga, D. (2021). A unified framework for closed-form nonparametric regression, classification, preference and mixed problems with skew gaussian processes. Machine Learning 110(11), 3095–3133.
A unified framework for closed-form nonparametric regression, classification, preference and mixed problems with skew gaussian processes
Authors: Benavoli, A. and Azzimonti, D. and Piga, D.
Year: 2021
Abstract: Skew-Gaussian Processes (SkewGPs) extend the multivariate Unified Skew-Normal distributions over finite dimensional vectors to distribution over functions. SkewGPs are more general and flexible than Gaussian processes, as SkewGPs may also represent asymmetric distributions. In a recent contribution, we showed that SkewGP and probit likelihood are conjugate, which allows us to compute the exact posterior for non-parametric binary classification and preference learning. In this paper, we generalize previous results and we prove that SkewGP is conjugate with both the normal and affine probit likelihood, and more in general, with their product. This allows us to (i) handle classification, preference, numeric and ordinal regression, and mixed problems in a unified framework; (ii) derive closed-form expression for the corresponding posterior distributions. We show empirically that the proposed framework based on SkewGP provides better performance than Gaussian processes in active learning and Bayesian (constrained) optimization. These two tasks are fundamental for design of experiments and in Data Science.
Published in Machine Learning 110(11), 3095–3133.
A unified framework for closed-form nonparametric regression, classification, preference and mixed problems with skew gaussian processes
@ARTICLE{azzimontid2021c,
title = {A unified framework for closed-form nonparametric regression, classification, preference and mixed problems with skew gaussian processes},
journal = {Machine Learning},
volume = {110},
author = {Benavoli, A. and Azzimonti, D. and Piga, D.},
number = {11},
pages = {3095--3133},
year = {2021},
doi = {10.1007/s10994-021-06039-x},
url = {}
}
DownloadBenavoli, A., Corani, G. (2021). State Space approximation of Gaussian Processes for time-series forecasting. In Proc. Workshop on Advanced Analytics and Learning on Temporal Data, 6th ECML PKDD Workshop, AALTD 2021.
State Space approximation of Gaussian Processes for time-series forecasting
Authors: Benavoli, A. and Corani, G.
Year: 2021
Abstract: Gaussian Processes (GPs), with a complex enough additive kernel, provide competitive results in time series forecasting compared to state-of-the-art approaches (arima, ETS) provided that: (i) during training the unnecessary components of the kernel are made irrelevant by automatic relevance determination; (ii) priors are assigned to each hyperparameter. However, GPs computational complexity grows cubically in time and quadratically in memory with the number of observations. The state space (SS) approximation of GPs allows to compute GPs based inferences with linear complexity. In this paper, we apply the SS representation to time series forecasting showing that SS models provide a performance comparable with that of full GP and better than state-of- the-art models (arima, ETS). Moreover, the SS representation allows us to derive new models by, for instance, combining ETS with kernels.
Published in Proc. Workshop on Advanced Analytics and Learning on Temporal Data, 6th ECML PKDD Workshop, AALTD 2021.
State Space approximation of Gaussian Processes for time-series forecasting
@INPROCEEDINGS{corani2021b,
title = {State {S}pace approximation of {G}aussian {P}rocesses for time-series forecasting},
booktitle = {Proc. Workshop on Advanced Analytics and Learning on Temporal Data, 6th {ECML} {PKDD} Workshop, {AALTD} 2021},
author = {Benavoli, A. and Corani, G.},
year = {2021},
doi = {},
url = {https://project.inria.fr/aaltd21/accepted-papers/}
}
DownloadBenavoli, A., Facchini, A., Zaffalon, M. (2021). The weirdness theorem and the origin of quantum paradoxes. Foundations of Physics 51(5), 95.
The weirdness theorem and the origin of quantum paradoxes
Authors: Benavoli, A., Facchini, A., Zaffalon, M.
Year: 2021
Abstract: We argue that there is a simple, unique, reason for all quantum paradoxes, and that such a reason is not uniquely related to quantum theory. It is rather a mathematical question that arises at the intersection of logic, probability, and computation. We give our ‘weirdness theorem’ that characterises the conditions under which the weirdness will show up. It shows that whenever logic has bounds due to the algorithmic nature of its tasks, then weirdness arises in the special form of negative probabilities or non-classical evaluation functionals. Weirdness is not logical inconsistency, however. It is only the expression of the clash between an unbounded and a bounded view of computation in logic. We discuss the implication of these results for quantum mechanics, arguing in particular that its interpretation should ultimately be computational rather than exclusively physical. We develop in addition a probabilistic theory in the real numbers that exhibits the phenomenon of entanglement, thus concretely showing that the latter is not specific to quantum mechanics.
Published in Foundations of Physics 51(5), 95.
The weirdness theorem and the origin of quantum paradoxes
@ARTICLE{benavoli2021b,
title = {The weirdness theorem and the origin of quantum paradoxes},
journal = {Foundations of Physics},
volume = {51},
author = {Benavoli, A., Facchini, A., Zaffalon, M.},
number = {5},
pages = {95},
year = {2021},
doi = {10.1007/s10701-021-00499-w},
url = {}
}
DownloadBenavoli, A., Facchini, A., Zaffalon, M. (2021). Quantum indistinguishability through exchangeable desirable gambles. In De Bock, J., Cano, A., Miranda, E., Moral, S. (Ed), ISIPTA 2021, PMLR 147, JMLR.org, pp. 22–31.
Quantum indistinguishability through exchangeable desirable gambles
Authors: Benavoli, A., Facchini, A., Zaffalon, M.
Year: 2021
Abstract: Two particles are identical if all their intrinsic properties, such as spin and charge, are the same, meaning that no quantum experiment can distinguish them. In addition to the well known principles of quantum mechanics, understanding systems of identical particles requires a new postulate, the so called symmetrization postulate}, which is thus crucial to the understanding of the physical properties of almost all kinds of aggregates of matter. In this work, we show that the postulate corresponds to exchangeability assessments for sets of observables (gambles) in a quantum experiment, when quantum mechanics is seen as a normative and algorithmic theory guiding an agent to assess her subjective beliefs represented as (coherent) sets of gambles. Finally, we show how sets of exchangeable observables (gambles) may be updated after a measurement and discuss the issue of defining entanglement for indistinguishable particle systems.
Accepted in De Bock, J., Cano, A., Miranda, E., Moral, S. (Ed), ISIPTA 2021, PMLR 147, JMLR.org, pp. 22–31.
Quantum indistinguishability through exchangeable desirable gambles
@INPROCEEDINGS{benavoli2021a,
title = {Quantum indistinguishability through exchangeable desirable gambles},
editor = {De Bock, J., Cano, A., Miranda, E., Moral, S.},
publisher = {JMLR.org},
series = {PMLR},
volume = {147},
booktitle = {{ISIPTA} 2021},
author = {Benavoli, A., Facchini, A., Zaffalon, M.},
pages = {22--31},
year = {2021},
doi = {},
url = {https://www.sipta.org/isipta21/pmlr/benavoli21.pdf}
}
DownloadBerjano, P., Langella, F., Ventriglia, L., Compagnone, D., Barletta, P., Huber, D., Mangili, F., Licandro, G., Galbusera, F., Cina, A., Bassani, T., Lamartina, C., Scaramuzzo, L., Bassani, R., Brayda-Bruno, M., Villafañe, J.H., Monti, L., Azzimonti, L. (2021). The influence of baseline clinical status and surgical strategy on early good to excellent result in spinal lumbar arthrodesis: a machine learning approach. Journal of Personalized Medicine 11(12).
The influence of baseline clinical status and surgical strategy on early good to excellent result in spinal lumbar arthrodesis: a machine learning approach
Authors: Berjano, P. and Langella, F. and Ventriglia, L. and Compagnone, D. and Barletta, P. and Huber, D. and Mangili, F. and Licandro, G. and Galbusera, F. and Cina, A. and Bassani, T. and Lamartina, C. and Scaramuzzo, L. and Bassani, R. and Brayda-Bruno, M. and Villafañe, J.H. and Monti, L. and Azzimonti, L.
Year: 2021
Abstract: The study aims to create a preoperative model from baseline demographic and health-related quality of life scores (HRQOL) to predict a good to excellent early clinical outcome using a machine learning (ML) approach. A single spine surgery center retrospective review of prospectively collected data from January 2016 to December 2020 from the institutional registry (SpineREG) was performed. The inclusion criteria were age ≥ 18 years, both sexes, lumbar arthrodesis procedure, a complete follow up assessment (Oswestry Disability Index—ODI, SF-36 and COMI back) and the capability to read and understand the Italian language. A delta of improvement of the ODI higher than 12.7/100 was considered a “good early outcome”. A combined target model of ODI (Δ ≥ 12.7/100), SF-36 PCS (Δ ≥ 6/100) and COMI back (Δ ≥ 2.2/10) was considered an “excellent early outcome”. The performance of the ML models was evaluated in terms of sensitivity, i.e., True Positive Rate (TPR), specificity, i.e., True Negative Rate (TNR), accuracy and area under the receiver operating characteristic curve (AUC ROC). A total of 1243 patients were included in this study. The model for predicting ODI at 6 months’ follow up showed a good balance between sensitivity (74.3%) and specificity (79.4%), while providing a good accuracy (75.8%) with ROC AUC = 0.842. The combined target model showed a sensitivity of 74.2% and specificity of 71.8%, with an accuracy of 72.8%, and an ROC AUC = 0.808. The results of our study suggest that a machine learning approach showed high performance in predicting early good to excellent clinical results.
Published in Journal of Personalized Medicine 11(12).
The influence of baseline clinical status and surgical strategy on early good to excellent result in spinal lumbar arthrodesis: a machine learning approach
@ARTICLE{azzimonti2021b,
title = {The influence of baseline clinical status and surgical strategy on early good to excellent result in spinal lumbar arthrodesis: a machine learning approach},
journal = {Journal of Personalized Medicine},
volume = {11},
author = {Berjano, P. and Langella, F. and Ventriglia, L. and Compagnone, D. and Barletta, P. and Huber, D. and Mangili, F. and Licandro, G. and Galbusera, F. and Cina, A. and Bassani, T. and Lamartina, C. and Scaramuzzo, L. and Bassani, R. and Brayda-Bruno, M. and Villafa\~ne, J.H. and Monti, L. and Azzimonti, L.},
number = {12},
year = {2021},
doi = {10.3390/jpm11121377},
url = {}
}
DownloadBianchi, F., Breschi, V., Piga, D., Piroddi, L. (2021). Model structure selection for switched narx system identification: a randomized approach. Automatica 125.
Model structure selection for switched narx system identification: a randomized approach
Authors: Bianchi, F. and Breschi, V. and Piga, D. and Piroddi, L.
Year: 2021
Abstract: The identification of switched systems is a challenging problem, which entails both combinatorial (sample-mode assignment) and continuous (parameter estimation) features. A general framework for this problem has been recently developed, which alternates between parameter estimation and sample-mode assignment, solving both tasks to global optimality under mild conditions. This article extends this framework to the nonlinear case, which further aggravates the combinatorial complexity of the identification problem, since a model structure selection task has to be addressed for each mode of the system. To solve this issue, we reformulate the learning problem in terms of the optimization of a probability distribution over the space of all possible model structures. Then, a randomized approach is employed to tune this distribution. The performance of the proposed approach on some benchmark examples is analyzed in detail.
Published in Automatica 125.
Model structure selection for switched narx system identification: a randomized approach
@ARTICLE{piga2021c,
title = {Model structure selection for switched narx system identification: a randomized approach},
journal = {Automatica},
volume = {125},
author = {Bianchi, F. and Breschi, V. and Piga, D. and Piroddi, L.},
year = {2021},
doi = {https://doi.org/10.1016/j.automatica.2020.109415},
url = {https://www.sciencedirect.com/science/article/pii/S0005109820306178}
}
DownloadBini, F., Pica, A., Azzimonti, L., Giusti, A., Ruinelli, L., Marinozzi, F., Trimboli, P. (2021). Artificial intelligence in thyroid field. A comprehensive review. Cancers 13(19).
Artificial intelligence in thyroid field. A comprehensive review
Authors: Bini, F. and Pica, A. and Azzimonti, L. and Giusti, A. and Ruinelli, L. and Marinozzi, F. and Trimboli, P.
Year: 2021
Abstract: Artificial intelligence (AI) uses mathematical algorithms to perform tasks that require human cognitive abilities. AI-based methodologies, e.g., machine learning and deep learning, as well as the recently developed research field of radiomics have noticeable potential to transform medical diagnostics. AI-based techniques applied to medical imaging allow to detect biological abnormalities, to diagnostic neoplasms or to predict the response to treatment. Nonetheless, the diagnostic accuracy of these methods is still a matter of debate. In this article, we first illustrate the key concepts and workflow characteristics of machine learning, deep learning and radiomics. We outline considerations regarding data input requirements, differences among these methodologies and their limitations. Subsequently, a concise overview is presented regarding the application of AI methods to the evaluation of thyroid images. We developed a critical discussion concerning limits and open challenges that should be addressed before the translation of AI techniques to the broad clinical use. Clarification of the pitfalls of AI-based techniques results crucial in order to ensure the optimal application for each patient.
Published in Cancers 13(19).
Artificial intelligence in thyroid field. A comprehensive review
@ARTICLE{azzimonti2021a,
title = {Artificial intelligence in thyroid field. A comprehensive review},
journal = {Cancers},
volume = {13},
author = {Bini, F. and Pica, A. and Azzimonti, L. and Giusti, A. and Ruinelli, L. and Marinozzi, F. and Trimboli, P.},
number = {19},
year = {2021},
doi = {10.3390/cancers13194740},
url = {https://www.mdpi.com/2072-6694/13/19/4740}
}
DownloadBodewes, T., Scutari, M. (2021). Learning bayesian networks from incomplete data with the node-averaged likelihood. International Journal of Approximate Reasoning 138, pp. 145–160.
Learning bayesian networks from incomplete data with the node-averaged likelihood
Authors: Bodewes, T. and Scutari, M.
Year: 2021
Published in International Journal of Approximate Reasoning 138, pp. 145–160.
Note: This is an extended version of the “Identifiability and Consistency of Bayesian Network Structure Learning from Incomplete Data” PMLR paper.
Learning bayesian networks from incomplete data with the node-averaged likelihood
@ARTICLE{scutari21d,
title = {Learning bayesian networks from incomplete data with the node-averaged likelihood},
journal = {International Journal of Approximate Reasoning},
volume = {138},
author = {Bodewes, T. and Scutari, M.},
pages = {145--160},
year = {2021},
doi = {},
url = {}
}
DownloadBonesana, C., Mangili, F., Antonucci, A. (2021). ADAPQUEST: a software for web-based adaptive questionnaires based on Bayesian networks. In AI4EDU: Artificial Intelligence for Education (@ Ijcai2021), Virtual Event.
ADAPQUEST: a software for web-based adaptive questionnaires based on Bayesian networks
Authors: Bonesana, C. and Mangili, F. and Antonucci, A.
Year: 2021
Abstract: We introduce ADAPQUEST, a software tool written in Java for the development of adaptive ques- tionnaires based on Bayesian networks. Adaptiveness is intended here as the dynamical choice of the question sequence on the basis of an evolving model of the skill level of the test taker. Bayesian networks offer a flexible and highly interpretable framework to describe such testing pro- cess, especially when coping with multiple skills. ADAPQUEST embeds dedicated elicitation strategies to simplify the elicitation of the questionnaire parameters. An application of this tool for the diag- nosis of mental disorders is also discussed together with some implementation details.
Published in AI4EDU: Artificial Intelligence for Education (@ Ijcai2021), Virtual Event.
Note: IJCAI2021 Workshop
ADAPQUEST: a software for web-based adaptive questionnaires based on Bayesian networks
@INPROCEEDINGS{bonesana2021a,
title = {{ADAPQUEST}: a software for web-based adaptive questionnaires based on {B}ayesian networks},
address = {Virtual Event},
booktitle = {{AI4EDU}: Artificial Intelligence for Education (@ Ijcai2021)},
author = {Bonesana, C. and Mangili, F. and Antonucci, A.},
year = {2021},
doi = {},
url = {https://arxiv.org/abs/2112.14476}
}
DownloadBregoli, A., Scutari, M., Stella, F. (2021). A constraint-based algorithm for the structural learning of continuous-time bayesian networks. International Journal of Approximate Reasoning 138, pp. 105–122.
A constraint-based algorithm for the structural learning of continuous-time bayesian networks
Authors: Bregoli, A. and Scutari, M. and Stella, F.
Year: 2021
Published in International Journal of Approximate Reasoning 138, pp. 105–122.
Note: This is an extended version of the “Constraint-Based Learning for Continuous-Time Bayesian Networks” PMLR paper.
A constraint-based algorithm for the structural learning of continuous-time bayesian networks
@ARTICLE{scutari21e,
title = {A constraint-based algorithm for the structural learning of continuous-time bayesian networks},
journal = {International Journal of Approximate Reasoning},
volume = {138},
author = {Bregoli, A. and Scutari, M. and Stella, F.},
pages = {105--122},
year = {2021},
doi = {},
url = {}
}
DownloadCabañas, R., Antonucci, A. (2021). CREPO: an open repository to benchmark credal network algorithms. In De Bock, J., Cano, A., Miranda, E., Moral, S. (Eds), International Symposium on Imprecise Probabilities: Theories and Applications (ISIPTA-2021) 147, JMLR.org.
CREPO: an open repository to benchmark credal network algorithms
Authors: Cabañas, R. and Antonucci, A.
Year: 2021
Abstract: Credal networks are a popular class of imprecise probabilistic graphical models obtained as a Bayesian network generalization based on, so-called credal, sets of probability mass functions. A Java library called CREMA has been recently released to model, process and query credal networks. Despite the NP-hardness of the (exact) task, a number of algorithms is available to approximate credal network inferences. In this paper we present CREPO, an open repository of synthetic credal networks, provided together with the exact results of inference tasks on these models. A Python tool is also delivered to load these data and interact with CREMA, thus making extremely easy to evaluate and compare existing and novel inference algorithms. To demonstrate such benchmarking scheme, we propose an approximate heuristic to be used inside variable elimination schemes to keep a bound on the maximum number of vertices generated during the combination step. A CREPO-based validation against approximate procedures based on linearization and exact techniques performed in CREMA is finally discussed.
Published in De Bock, J., Cano, A., Miranda, E., Moral, S. (Eds), International Symposium on Imprecise Probabilities: Theories and Applications (ISIPTA-2021) 147, JMLR.org.
CREPO: an open repository to benchmark credal network algorithms
@INPROCEEDINGS{cabanas2021b,
title = {{CREPO}: an open repository to benchmark credal network algorithms},
editor = {De Bock, J. and Cano, A. and Miranda, E. and Moral, S.},
publisher = {JMLR.org},
volume = {147},
booktitle = {International Symposium on Imprecise Probabilities: Theories and Applications ({ISIPTA}-2021)},
author = {Caba\~nas, R. and Antonucci, A.},
year = {2021},
doi = {},
url = {http://proceedings.mlr.press/v147/cabanas21a/cabanas21a.pdf}
}
DownloadCasanova, A., Benavoli, A., Zaffalon, M. (2021). Nonlinear desirability as a linear classification problem. In Proceedings of Machine Learning Research 147, pp. 61-71.
Nonlinear desirability as a linear classification problem
Authors: Casanova, A. and Benavoli, A. and Zaffalon, M.
Year: 2021
Abstract: The present paper proposes a generalization of linearity
axioms of coherence through a geometrical approach,
which leads to an alternative interpretation of
desirability as a classification problem. In particular,
we analyze different sets of rationality axioms and,
for each one of them, we show that proving that a
subject, who provides finite accept and reject statements,
respects these axioms, corresponds to solving
a binary classification task using, each time, a different
(usually nonlinear) family of classifiers. Moreover,
by borrowing ideas from machine learning, we show
the possibility to define a feature mapping allowing
us to reformulate the above nonlinear classification
problems as linear ones in a higher-dimensional space.
This allows us to interpret gambles directly as payoffs
vectors of monetary lotteries, as well as to reduce the task of proving the rationality of a subject to a linear
classification task.
Published in Proceedings of Machine Learning Research 147, pp. 61-71.
Nonlinear desirability as a linear classification problem
@INPROCEEDINGS{casanova2021c,
title = {Nonlinear desirability as a linear classification problem},
volume = {147},
booktitle = {Proceedings of Machine Learning Research},
author = {Casanova, A. and Benavoli, A. and Zaffalon, M.},
pages = {61-71},
year = {2021},
doi = {},
url = {https://proceedings.mlr.press/v147/casanova21a.html}
}
DownloadCasanova, A., Kohlas, J., Zaffalon, M. (2021). Algebras of sets and coherent sets of gambles. In European Conference on Symbolic and Quantitative Approaches with Uncertainty, pp. 603-615.
Algebras of sets and coherent sets of gambles
Authors: Casanova, A. and Kohlas, J. and Zaffalon, M.
Year: 2021
Abstract: In a recent work we have shown how to construct an information algebra of coherent sets of gambles defined on general possibility spaces. Here we analyze the connection of such an algebra with the set algebra of sets of its atoms and the set algebra of subsets of the possibility space on which gambles are defined. Set algebras are particularly important information algebras since they are their prototypical structures. Furthermore, they are the algebraic counterparts of classical propositional logic. As a consequence, this paper also details how propositional logic is naturally embedded into the theory of imprecise probabilities.
Published in European Conference on Symbolic and Quantitative Approaches with Uncertainty, pp. 603-615.
Algebras of sets and coherent sets of gambles
@INPROCEEDINGS{casanova2021d,
title = {Algebras of sets and coherent sets of gambles},
booktitle = {European Conference on Symbolic and Quantitative Approaches {w}ith Uncertainty},
author = {Casanova, A. and Kohlas, J. and Zaffalon, M.},
pages = {603-615},
year = {2021},
doi = {10.1007/978-3-030-86772-0_43},
url = {}
}
DownloadCasanova, A., Miranda, E., Zaffalon, M. (2021). Joint desirability foundations of social choice and opinion pooling. Annals of Mathematics and Artificial Intelligence 89(10–11), pp. 965–1011.
Joint desirability foundations of social choice and opinion pooling
Authors: Casanova, A. and Miranda, E. and Zaffalon, M.
Year: 2021
Abstract: We develop joint foundations for the fields of social choice and opinion pooling using coherent sets of desirable gambles, a general uncertainty model that allows to encompass both complete and incomplete preferences. This leads on the one hand to a new perspective of traditional results of social choice (in particular Arrow’s theorem as well as sufficient conditions for the existence of an oligarchy and democracy) and on the other hand to using the same framework to analyse opinion pooling. In particular, we argue that weak Pareto (unanimity) should be given the status of a rationality requirement and use this to discuss the aggregation of experts’ opinions based on probability and (state-independent) utility, showing some inherent limitation of this framework, with implications for statistics. The connection between our results and earlier work in the literature is also discussed.
Published in Annals of Mathematics and Artificial Intelligence 89(10–11), pp. 965–1011.
Joint desirability foundations of social choice and opinion pooling
@ARTICLE{casanova2021a,
title = {Joint desirability foundations of social choice and opinion pooling},
journal = {Annals of Mathematics and Artificial Intelligence},
volume = {89},
author = {Casanova, A. and Miranda, E. and Zaffalon, M.},
number = {10--11},
pages = {965--1011},
year = {2021},
doi = {10.1007/s10472-021-09733-7},
url = {}
}
DownloadCorani, G., Benavoli, A., Zaffalon, M. (2021). Time series forecasting with Gaussian Processes needs priors. In European Conference on Machine Learning and Knowledge Discovery in Databases, pp. 103–117.
Time series forecasting with Gaussian Processes needs priors
Authors: Corani, G. and Benavoli, A. and Zaffalon, M.
Year: 2021
Abstract: Automatic forecasting is the task of receiving a time series and returning a forecast for the next time steps without any human intervention. Gaussian Processes (GPs) are a powerful tool for modeling time series, but so far there are no competitive approaches for automatic forecasting based on GPs. We propose practical solutions to two problems: automatic selection of the optimal kernel and reliable estimation of the hyperparameters.
We propose a fixed composition of kernels, which contains the components needed to model most time series: linear trend, periodic patterns, and other flexible kernel for modeling the non-linear trend. Not all components are necessary to model each time series; during training the unnecessary components are automatically made irrelevant via automatic relevance determination (ARD). We moreover assign priors to the hyperparameters, in order to keep the inference within a plausible range; we design such priors through an empirical Bayes approach. We present results on many time series of different types; our GP model is more accurate than state-of-the-art time series models. Thanks to the priors, a single restart is enough the estimate the hyperparameters; hence the model is also fast to train.
Published in European Conference on Machine Learning and Knowledge Discovery in Databases, pp. 103–117.
Time series forecasting with Gaussian Processes needs priors
@INPROCEEDINGS{corani2021a,
title = {Time series forecasting with {G}aussian {P}rocesses needs priors},
booktitle = {European Conference on Machine Learning and Knowledge Discovery in Databases},
author = {Corani, G. and Benavoli, A. and Zaffalon, M.},
pages = {103--117},
year = {2021},
doi = {10.1007/978-3-030-86514-6_7},
url = {}
}
DownloadForgione, M., Piga, D. (2021). dynoNet: a neural network architecture for learning dynamical systems. International Journal of Adaptive Control and Signal Processing 35, pp. 612–626.
dynoNet: a neural network architecture for learning dynamical systems
Authors: Forgione, M. and Piga, D.
Year: 2021
Abstract: This paper introduces a network architecture, called dynoNet, utilizing linear dynamical operators as elementary building blocks. Owing to the dynamical nature of these blocks, dynoNet networks are tailored for sequence modeling and system identification purposes. The backpropagation behavior of the linear dynamical operator with respect to both its parameters and its input sequence is defined. This enables end-toend training of structured networks containing linear dynamical operators and other differentiable units, exploiting existing deep learning software. Examples show the effectiveness of the proposed approach on well-known system identification benchmarks.
Published in International Journal of Adaptive Control and Signal Processing 35, pp. 612–626.
dynoNet: a neural network architecture for learning dynamical systems
@ARTICLE{forgione2021a,
title = {{dynoNet}: a neural network architecture for learning dynamical systems},
journal = {International Journal of Adaptive Control and Signal Processing},
volume = {35},
author = {Forgione, M. and Piga, D.},
pages = {612--626},
year = {2021},
doi = {10.1002/acs.3216},
url = {}
}
DownloadForgione, M., Piga, D. (2021). Continuous-time system identification with neural networks: model structures and fitting criteria. European Journal of Control 59, pp. 69–81.
Continuous-time system identification with neural networks: model structures and fitting criteria
Authors: Forgione, M. and Piga, D.
Year: 2021
Abstract: This paper presents tailor-made neural model structures and two custom fitting criteria for learning dynamical systems. The proposed framework is based on a representation of the system behavior in terms of continuous-time state-space models. The sequence of hidden states is optimized along with the neural network parameters in order to minimize the difference between measured and estimated outputs, and at the same time to guarantee that the optimized state sequence is consistent with the estimated system dynamics. The effectiveness of the approach is demonstrated through three case studies, including two public system identification benchmarks based on experimental data.
Published in European Journal of Control 59, pp. 69–81.
Continuous-time system identification with neural networks: model structures and fitting criteria
@ARTICLE{forgione2021b,
title = {Continuous-time system identification with neural networks: model structures and fitting criteria},
journal = {European Journal of Control},
volume = {59},
author = {Forgione, M. and Piga, D.},
pages = {69--81},
year = {2021},
doi = {10.1016/j.ejcon.2021.01.008},
url = {}
}
DownloadDalla Gasperina, S., Roveda, L., Pedrocchi, A., Braghin, F., Gandolla, M. (2021). Review on patient-cooperative control strategies for upper-limb rehabilitation exoskeletons. Frontiers in Robotics and AI.
Review on patient-cooperative control strategies for upper-limb rehabilitation exoskeletons
Authors: Dalla Gasperina, S. and Roveda, L. and Pedrocchi, A. and Braghin, F. and Gandolla, M.
Year: 2021
Abstract: Technology-supported rehabilitation therapy for neurological patients has gained increasing interest since the last decades. The literature agrees that the goal of robots should be to induce motor plasticity in subjects undergoing rehabilitation treatment by providing the patients with repetitive, intensive, and task-oriented treatment. As a key element, robot controllers should adapt to patients’ status and recovery stage. Thus, the design of effective training modalities and their hardware implementation play a crucial role in robot-assisted rehabilitation and strongly influence the treatment outcome. The objective of this paper is to provide a multi-disciplinary vision of patient-cooperative control strategies for upper-limb rehabilitation exoskeletons to help researchers bridge the gap between human motor control aspects, desired rehabilitation training modalities, and their hardware implementations. To this aim, we propose a three-level classification based on i) ”high-level” training modalities, ii) ”low-level” control strategies, and iii) ”hardware-level” implementation. Then, we provide examples of literature upper-limb exoskeletons to show how the three levels of implementation have been combined to obtain a given high-level behavior, which is specifically designed to promote motor relearning during the rehabilitation treatment. Finally, we emphasize the need for the development of compliant control strategies, based on the collaboration between the exoskeleton and the wearer, we report the key findings to promote the desired physical human-robot interaction for neurorehabilitation, and we provide insights and suggestions for future works.
Published in Frontiers in Robotics and AI.
Review on patient-cooperative control strategies for upper-limb rehabilitation exoskeletons
@ARTICLE{Roveda2021e,
title = {Review on patient-cooperative control strategies for upper-limb rehabilitation exoskeletons},
journal = {Frontiers in Robotics and {AI}},
author = {Dalla Gasperina, S. and Roveda, L. and Pedrocchi, A. and Braghin, F. and Gandolla, M.},
year = {2021},
doi = {https://www.frontiersin.org/articles/10.3389/frobt.2021.745018/abstract},
url = {}
}
DownloadGeluykens, J., Mitrovic, S., Vázquez, C.O., Laino, T., Vaucher, A.C., Weerdt, J.D. (2021). Neural machine translation for conditional generation of novel procedures. In 54th Hawaii International Conference on System Sciences, HICSS 2021, Kauai, Hawaii, Usa, January 5, 2021, ScholarSpace, pp. 1–10.
Neural machine translation for conditional generation of novel procedures
Authors: Geluykens, J. and Mitrovic, S. and Vázquez, C.O. and Laino, T. and Vaucher, A.C. and Weerdt, J.D.
Year: 2021
Published in 54th Hawaii International Conference on System Sciences, HICSS 2021, Kauai, Hawaii, Usa, January 5, 2021, ScholarSpace, pp. 1–10.
Neural machine translation for conditional generation of novel procedures
@INPROCEEDINGS{mitrovic2021b,
title = {Neural machine translation for conditional generation of novel procedures},
publisher = {ScholarSpace},
booktitle = {54th Hawaii International Conference on System Sciences, {HICSS} 2021, Kauai, Hawaii, Usa, January 5, 2021},
author = {Geluykens, J. and Mitrovic, S. and V\'azquez, C.O. and Laino, T. and Vaucher, A.C. and Weerdt, J.D.},
pages = {1--10},
year = {2021},
doi = {},
url = {http://hdl.handle.net/10125/70744}
}
DownloadGómez-Olmedo, M., Cabañas, R., Cano, A., Moral, S., Retamero, O.P. (2021). Value-based potentials: exploiting quantitative information regularity patterns in probabilistic graphical models. International Journal of Intelligent Systems.
Value-based potentials: exploiting quantitative information regularity patterns in probabilistic graphical models
Authors: Gómez-Olmedo, M. and Cabañas, R. and Cano, A. and Moral, S. and Retamero, O.P.
Year: 2021
Published in International Journal of Intelligent Systems.
Value-based potentials: exploiting quantitative information regularity patterns in probabilistic graphical models
@ARTICLE{cabanas2021c,
title = {Value-based potentials: exploiting quantitative information regularity patterns in probabilistic graphical models},
journal = {International Journal of Intelligent Systems},
author = {G\'omez-Olmedo, M. and Caba\~nas, R. and Cano, A. and Moral, S. and Retamero, O.P.},
year = {2021},
doi = {},
url = {}
}
DownloadGrasso, G., Gregorio, A.D., Mavkov, B., Piga, D., Labate, G.F.D., Danani, A., Deriu, M.A. (2021). Fragmented blind docking: a novel protein–ligand binding prediction protocol. Journal of Biomolecular Structure and Dynamics.
Fragmented blind docking: a novel protein–ligand binding prediction protocol
Authors: Grasso, G. and Gregorio, A.D. and Mavkov, B. and Piga, D. and Labate, G.F.D. and Danani, A. and Deriu, M.A.
Year: 2021
Abstract: AbstractIn the present paper we propose a novel blind docking protocol based on Autodock-Vina. The developed docking protocol can provide binding site identification and binding pose prediction at the same time, by a systematical exploration of the protein volume performed with several preliminary docking calculations. In our opinion, this protocol can be successfully applied during the first steps of the virtual screening pipeline, because it provides binding site identification and binding pose prediction at the same time without visual evaluation of the binding site. After the binding pose prediction, MM/GBSA re-scoring rescoring procedures has been applied to improve the accuracy of the protein–ligand bound state. The FRAD protocol has been tested on 116 protein–ligand complexes of the Heat Shock Protein 90 – alpha, on 176 of Human Immunodeficiency virus protease 1, and on more than 100 protein–ligand system taken from the PDBbind dataset. Overall, the FRAD approach combined to MM/GBSA re-scoring can be considered as a powerful tool to increase the accuracy and efficiency with respect to other standard docking approaches when the ligand-binding site is unknown.Communicated by Ramaswamy H. Sarma
Published in Journal of Biomolecular Structure and Dynamics.
Fragmented blind docking: a novel protein–ligand binding prediction protocol
@ARTICLE{piga2021f,
title = {Fragmented blind docking: a novel protein--ligand binding prediction protocol},
journal = {Journal of Biomolecular Structure and Dynamics},
author = {Grasso, G. and Gregorio, A.D. and Mavkov, B. and Piga, D. and Labate, G.F.D. and Danani, A. and Deriu, M.A.},
year = {2021},
doi = {10.1080/07391102.2021.1988709},
url = {https://doi.org/10.1080/07391102.2021.1988709}
}
DownloadHalasz, G., Sperti, M., Villani, M., Michelucci, U., Agostoni, P., Biagi, A., Rossi, L., Botti, A., Mari, C., Maccarini, M., Pura, F., Roveda, L., Nardecchia, A., Mottola, E., Nolli, M., Salvioni, E., Mapelli, M., Deriu, M.A., Piga, D., Piepoli, M. (2021). A machine learning approach for mortality prediction in covid-19 pneumonia: development and evaluation of the piacenza score. Journal of Medical Internet Research 23(5).
A machine learning approach for mortality prediction in covid-19 pneumonia: development and evaluation of the piacenza score
Authors: Halasz, G. and Sperti, M. and Villani, M. and Michelucci, U. and Agostoni, P. and Biagi, A. and Rossi, L. and Botti, A. and Mari, C. and Maccarini, M. and Pura, F. and Roveda, L. and Nardecchia, A. and Mottola, E. and Nolli, M. and Salvioni, E. and Mapelli, M. and Deriu, M.A. and Piga, D. and Piepoli, M.
Year: 2021
Abstract: Background: Several models have been developed to predict mortality in patients with COVID-19 pneumonia, but only a few have demonstrated enough discriminatory capacity. Machine learning algorithms represent a novel approach for the data-driven prediction of clinical outcomes with advantages over statistical modeling. Objective: We aimed to develop a machine learning–based score–-the Piacenza score–-for 30-day mortality prediction in patients with COVID-19 pneumonia. Methods: The study comprised 852 patients with COVID-19 pneumonia, admitted to the Guglielmo da Saliceto Hospital in Italy from February to November 2020. Patients' medical history, demographics, and clinical data were collected using an electronic health record. The overall patient data set was randomly split into derivation and test cohorts. The score was obtained through the naïve Bayes classifier and externally validated on 86 patients admitted to Centro Cardiologico Monzino (Italy) in February 2020. Using a forward-search algorithm, 6 features were identified: age, mean corpuscular hemoglobin concentration, PaO2/FiO2 ratio, temperature, previous stroke, and gender. The Brier index was used to evaluate the ability of the machine learning model to stratify and predict the observed outcomes. A user-friendly website was designed and developed to enable fast and easy use of the tool by physicians. Regarding the customization properties of the Piacenza score, we added a tailored version of the algorithm to the website, which enables an optimized computation of the mortality risk score for a patient when some of the variables used by the Piacenza score are not available. In this case, the naïve Bayes classifier is retrained over the same derivation cohort but using a different set of patient characteristics. We also compared the Piacenza score with the 4C score and with a naïve Bayes algorithm with 14 features chosen a priori. Results: The Piacenza score exhibited an area under the receiver operating characteristic curve (AUC) of 0.78 (95% CI 0.74-0.84, Brier score=0.19) in the internal validation cohort and 0.79 (95% CI 0.68-0.89, Brier score=0.16) in the external validation cohort, showing a comparable accuracy with respect to the 4C score and to the naïve Bayes model with a priori chosen features; this achieved an AUC of 0.78 (95% CI 0.73-0.83, Brier score=0.26) and 0.80 (95% CI 0.75-0.86, Brier score=0.17), respectively. Conclusions: Our findings demonstrated that a customizable machine learning–based score with a purely data-driven selection of features is feasible and effective for the prediction of mortality among patients with COVID-19 pneumonia.
Published in Journal of Medical Internet Research 23(5).
A machine learning approach for mortality prediction in covid-19 pneumonia: development and evaluation of the piacenza score
@ARTICLE{piga2021g,
title = {A machine learning approach for mortality prediction in covid-19 pneumonia: development and evaluation of the piacenza score},
journal = {Journal of Medical Internet Research},
volume = {23},
author = {Halasz, G. and Sperti, M. and Villani, M. and Michelucci, U. and Agostoni, P. and Biagi, A. and Rossi, L. and Botti, A. and Mari, C. and Maccarini, M. and Pura, F. and Roveda, L. and Nardecchia, A. and Mottola, E. and Nolli, M. and Salvioni, E. and Mapelli, M. and Deriu, M.A. and Piga, D. and Piepoli, M.},
number = {5},
year = {2021},
doi = {10.2196/29058},
url = {https://www.jmir.org/2021/5/e29058}
}
DownloadKania, L., Schürch, M., Azzimonti, D., Benavoli, A. (2021). Sparse information filter for fast gaussian process regression. In Oliver, N., Pérez-Cruz, F., Kramer, S., Read, J., Lozano, J.A. (Eds), Machine Learning and Knowledge Discovery in Databases. Research Track, Springer International Publishing, Cham, pp. 527–542.
Sparse information filter for fast gaussian process regression
Authors: Kania, L. and Schürch, M. and Azzimonti, D. and Benavoli, A.
Year: 2021
Abstract: Gaussian processes (GPs) are an important tool in machine learning and applied mathematics with applications ranging from Bayesian optimization to calibration of computer experiments. They constitute a powerful kernelized non-parametric method with well-calibrated uncertainty estimates, however, off-the-shelf GP inference procedures are limited to datasets with a few thousand data points because of their cubic computational complexity. For this reason, many sparse GPs techniques were developed over the past years. In this paper, we focus on GP regression tasks and propose a new algorithm to train variational sparse GP models. An analytical posterior update expression based on the Information Filter is derived for the variational sparse GP model. We benchmark our method on several real datasets with millions of data points against the state-of-the-art Stochastic Variational GP (SVGP) and sparse orthogonal variational inference for Gaussian Processes (SOLVEGP). Our method achieves comparable performances to SVGP and SOLVEGP while providing considerable speed-ups. Specifically, it is consistently four times faster than SVGP and on average 2.5 times faster than SOLVEGP.
Published in Oliver, N., Pérez-Cruz, F., Kramer, S., Read, J., Lozano, J.A. (Eds), Machine Learning and Knowledge Discovery in Databases. Research Track, Springer International Publishing, Cham, pp. 527–542.
Sparse information filter for fast gaussian process regression
@INPROCEEDINGS{schurch2021a,
title = {Sparse information filter for fast gaussian process regression},
editor = {Oliver, N. and P\'erez-Cruz, F. and Kramer, S. and Read, J. and Lozano, J.A.},
publisher = {Springer International Publishing},
address = {Cham},
booktitle = {Machine Learning and Knowledge Discovery in Databases. Research Track},
author = {Kania, L. and Sch\"urch, M. and Azzimonti, D. and Benavoli, A.},
pages = {527--542},
year = {2021},
doi = {10.1007/978-3-030-86523-8_32},
url = {}
}
DownloadKanjirangat, V., Rinaldi, F. (2021). Enhancing Biomedical Relation Extraction with Transformer Models using Shortest Dependency Path Features and Triplet Information. Journal of Biomedical Informatics.
Enhancing Biomedical Relation Extraction with Transformer Models using Shortest Dependency Path Features and Triplet Information
Authors: Kanjirangat, V. and Rinaldi, F.
Year: 2021
Abstract: Entity relation extraction plays an important role in the biomedical, healthcare, and clinical research areas. Recently, pre-trained models based on transformer architectures and their variants have shown remarkable performances in various natural language processing tasks. Most of these variants were based on slight modifications in the architectural components, representation schemes and augmenting data using distant supervision methods. In distantly supervised methods, one of the main challenges is pruning out noisy samples. A similar situation can arise when the training samples are not directly available but need to be constructed from the given dataset. The BioCreative V Chemical Disease Relation (CDR) task provides a dataset that does not explicitly offer mention-level gold annotations and hence replicates the above scenario. Selecting the representative sentences from the given abstract or document text that could convey a potential entity relationship becomes essential. Most of the existing methods in literature propose to either consider the entire text or all the sentences which contain the entity mentions. This could be a computationally expensive and time consuming approach. This paper presents a novel approach to handle such scenarios, specifically in biomedical relation extraction. We propose utilizing the Shortest Dependency Path (SDP) features for constructing data samples by pruning out noisy information and selecting the most representative samples for model learning. We also utilize triplet information in model learning using the biomedical variant of BERT, viz., BioBERT. The problem is represented as a sentence pair classification task using the sentence and the entity-relation pair as input. We analyze the approach on both intra-sentential and inter-sentential relations in the CDR dataset. The proposed approach that utilizes the SDP and triplet features presents promising results, specifically on the inter-sentential relation extraction task.
Published in Journal of Biomedical Informatics, Elsevier.
Enhancing Biomedical Relation Extraction with Transformer Models using Shortest Dependency Path Features and Triplet Information
@ARTICLE{vani2021b,
title = {Enhancing {B}iomedical {R}elation {E}xtraction with {T}ransformer {M}odels using {S}hortest {D}ependency {P}ath {F}eatures and {T}riplet {I}nformation},
journal = {Journal of Biomedical Informatics},
publisher = {Elsevier},
author = {Kanjirangat, V. and Rinaldi, F.},
year = {2021},
doi = {10.1016/j.jbi.2021.103893},
url = {https://www.sciencedirect.com/science/article/pii/S1532046421002227}
}
DownloadKohlas, J., Casanova, A., Zaffalon, M. (2021). Information algebras of coherent sets of gambles in general possibility spaces. In Proceedings of Machine Learning Research 147, 191–200.
Information algebras of coherent sets of gambles in general possibility spaces
Authors: Kohlas, J. and Casanova, A. and Zaffalon, M.
Year: 2021
Abstract: In this paper, we show that coherent sets of gambles
can be embedded into the algebraic structure of information algebra. This leads firstly, to a new perspective
of the algebraic and logical structure of desirability
and secondly, it connects desirability, hence imprecise probabilities, to other formalism in computer science sharing the same underlying structure. Both the
domain-free and the labeled view of the information
algebra of coherent sets of gambles are presented, considering general possibility spaces.
Published in Proceedings of Machine Learning Research 147, 191–200.
Information algebras of coherent sets of gambles in general possibility spaces
@INPROCEEDINGS{casanova2021b,
title = {Information algebras of coherent sets of gambles in general possibility spaces},
volume = {147},
booktitle = {Proceedings of Machine Learning Research},
author = {Kohlas, J. and Casanova, A. and Zaffalon, M.},
pages = {191--200},
year = {2021},
doi = {},
url = {http://proceedings.mlr.press/v147/kohlas21a/kohlas21a.pdf}
}
DownloadLiew, B.X.W., Ford, J.J., Scutari, M., Hahne, A.J. (2021). Using data-driven bayesian network analysis to explore recovery pathways in people with low back pain receiving individualised physiotherapy or advice. PLoS ONE 16(10), e0258515.
Using data-driven bayesian network analysis to explore recovery pathways in people with low back pain receiving individualised physiotherapy or advice
Authors: Liew, B.X.W. and Ford, J.J. and Scutari, M. and Hahne, A.J.
Year: 2021
Published in PLoS ONE 16(10), e0258515.
Using data-driven bayesian network analysis to explore recovery pathways in people with low back pain receiving individualised physiotherapy or advice
@ARTICLE{scutari21c,
title = {Using data-driven bayesian network analysis to explore recovery pathways in people with low back pain receiving individualised physiotherapy or advice},
journal = {{PLoS} {ONE}},
volume = {16},
author = {Liew, B.X.W. and Ford, J.J. and Scutari, M. and Hahne, A.J.},
number = {10},
pages = {e0258515},
year = {2021},
doi = {},
url = {}
}
DownloadLiew, B.X.W., Peolsson, A., Falla, D., Cleland, J.A., Scutari, M., Kierkegaard, M., r A Dedering, (2021). Mechanisms of recovery after neck-specific or general exercises in patients with cervical radiculopathy. European Journal of Pain 25(5), pp. 1162–1172.
Mechanisms of recovery after neck-specific or general exercises in patients with cervical radiculopathy
Authors: Liew, B.X.W. and Peolsson, A. and Falla, D. and Cleland, J.A. and Scutari, M. and Kierkegaard, M. and r A Dedering,
Year: 2021
Published in European Journal of Pain 25(5), pp. 1162–1172.
Mechanisms of recovery after neck-specific or general exercises in patients with cervical radiculopathy
@ARTICLE{scutari21b,
title = {Mechanisms of recovery after neck-specific or general exercises in patients with cervical radiculopathy},
journal = {European Journal of Pain},
volume = {25},
author = {Liew, B.X.W. and Peolsson, A. and Falla, D. and Cleland, J.A. and Scutari, M. and Kierkegaard, M. and r A Dedering, },
number = {5},
pages = {1162--1172},
year = {2021},
doi = {},
url = {}
}
DownloadLlerena, J.V., Mauá, D.D., Antonucci, A. (2021). Cautious classification with data missing not at random using generative random forests. In Vejnarová, J., Wilson, N. (Eds), Symbolic and Quantitative Approaches to Reasoning With Uncertainty, Springer International Publishing, Cham, pp. 284–298.
Cautious classification with data missing not at random using generative random forests
Authors: Llerena, J.V. and Mauá, D.D. and Antonucci, A.
Year: 2021
Abstract: Missing data present a challenge for most machine learning approaches. When a generative probabilistic model of the data is available, an effective approach is to marginalize missing values out. Probabilistic circuits are expressive generative models that allow for efficient exact inference. However, data is often missing not at random, and marginalization can lead to overconfident and wrong conclusions. In this work, we develop an efficient algorithm for assessing the robustness of classifications made by probabilistic circuits to imputations of the non-ignorable portion of missing data at prediction time. We show that our algorithm is exact when the model satisfies certain constraints, which is the case for the recent proposed Generative Random Forests, that equip Random Forest Classifiers with a full probabilistic model of the data. We also show how to extend our approach to handle non-ignorable missing data at training time.
Published in Vejnarová, J., Wilson, N. (Eds), Symbolic and Quantitative Approaches to Reasoning With Uncertainty, Springer International Publishing, Cham, pp. 284–298.
Cautious classification with data missing not at random using generative random forests
@INPROCEEDINGS{antonucci2021b,
title = {Cautious classification with data missing not at random using generative random forests},
editor = {Vejnarov\'a, J. and Wilson, N.},
publisher = {Springer International Publishing},
address = {Cham},
booktitle = {Symbolic and Quantitative Approaches to Reasoning With Uncertainty},
author = {Llerena, J.V. and Mau\'a, D.D. and Antonucci, A.},
pages = {284--298},
year = {2021},
doi = {10.1007/978-3-030-86772-0_21},
url = {}
}
DownloadMaggioni, B., Marescotti, E., Zanchettin, M., Piga, D., Roveda, L. (2021). Velocity planning of a robotic task enhanced by fuzzy logic and dynamic movement primitives . In .
Velocity planning of a robotic task enhanced by fuzzy logic and dynamic movement primitives
Authors: Maggioni, B. and Marescotti, E. and Zanchettin, M. and Piga, D. and Roveda, L.
Year: 2021
Abstract: Many industrial tasks, such as welding and sealing, require not only a precise path reference, but also an advanced
velocity planning in order to achieve the target quality for the final products. In this paper, a novel approach is proposed to
perform robotic trajectory planning. The developed algorithm exploits Fuzzy Logic (FL) to relate the path features (such as
curves or sharp edges) to the proper execution velocity. Such a computed velocity reference is then used as an input for
Dynamical Movement Primitives (DMP), providing the reference signals to the robot controller. The main improved
methodology features are: path-based velocity planning, extension of DMP to variable velocity reference and smoothing of
the velocity reference including robot velocity/acceleration limits. The algorithm can be implemented in a collaborative
framework, defining a compliant controller embedded into the DMP for online trajectory planning.
Published in ARCI 2021.
Velocity planning of a robotic task enhanced by fuzzy logic and dynamic movement primitives
@INPROCEEDINGS{Roveda2021g,
title = {Velocity planning of a robotic task enhanced by fuzzy logic and dynamic movement primitives },
journal = {{ARCI} 2021},
author = {Maggioni, B. and Marescotti, E. and Zanchettin, M. and Piga, D. and Roveda, L.},
year = {2021},
doi = {https://www.sensorsportal.com/ARCI/ARCI_2021_Proceedings.pdf#page=26},
url = {}
}
DownloadMasegosa, A.R., Cabañas, R., Langseth, H., Nielsen, T.D., Salmerón, A. (2021). Probabilistic models with deep neural networks. Entropy 23(1).
Probabilistic models with deep neural networks
Authors: Masegosa, A.R. and Cabañas, R. and Langseth, H. and Nielsen, T.D. and Salmerón, A.
Year: 2021
Abstract: Recent advances in statistical inference have significantly expanded the toolbox of probabilistic modeling. Historically, probabilistic modeling has been constrained to very restricted model classes, where exact or approximate probabilistic inference is feasible. However, developments in variational inference, a general form of approximate probabilistic inference that originated in statistical physics, have enabled probabilistic modeling to overcome these limitations: (i) Approximate probabilistic inference is now possible over a broad class of probabilistic models containing a large number of parameters, and (ii) scalable inference methods based on stochastic gradient descent and distributed computing engines allow probabilistic modeling to be applied to massive data sets. One important practical consequence of these advances is the possibility to include deep neural networks within probabilistic models, thereby capturing complex non-linear stochastic relationships between the random variables. These advances, in conjunction with the release of novel probabilistic modeling toolboxes, have greatly expanded the scope of applications of probabilistic models, and allowed the models to take advantage of the recent strides made by the deep learning community. In this paper, we provide an overview of the main concepts, methods, and tools needed to use deep neural networks within a probabilistic modeling framework.
Published in Entropy 23(1).
Probabilistic models with deep neural networks
@ARTICLE{cabanas2021a,
title = {Probabilistic models with deep neural networks},
journal = {Entropy},
volume = {23},
author = {Masegosa, A.R. and Caba\~nas, R. and Langseth, H. and Nielsen, T.D. and Salmer\'on, A.},
number = {1},
year = {2021},
doi = {10.3390/e23010117},
url = {https://www.mdpi.com/1099-4300/23/1/117}
}
DownloadMejari, M., Mavkov, B., Forgione, M., Piga, D. (2021). An integral architecture for identification of continuous-time state-space lpv models. In 4th IFAC Workshop on Linear Parameter-Varying Systems LPVS 2021, Milan, Italy.
An integral architecture for identification of continuous-time state-space lpv models
Authors: Mejari, M. and Mavkov, B. and Forgione, M. and Piga, D.
Year: 2021
Abstract: This paper presents an integral architecture for direct identifi cation of continuous-time linear parameter-varying (LPV) state-space models. The main building block of the
proposed architecture consist of an LPV model followed by an integral block, which is used to approximate the continuous-time state map of an LPV representation. The unknown LPV model matrices are estimated along with the state sequence by minimizing a properly constructed dual-objective criterion. A coordinate descent algorithm is employed to optimize the desired objective, which alternates between computing the unknown LPV matrices and estimating the state sequence. Thanks to the linear parametric structure induced by the LPV models, the unknown parameters within each coordinate descent step can be computed analytically via ordinary least squares. The effectiveness of the proposed methodology is assessed via a numerical example.
Published in 4th IFAC Workshop on Linear Parameter-Varying Systems LPVS 2021, Milan, Italy.
An integral architecture for identification of continuous-time state-space lpv models
@INPROCEEDINGS{mejari2021a,
title = {An integral architecture for identification of continuous-time state-space lpv models},
address = {Milan, Italy},
booktitle = {4th {IFAC} Workshop on Linear Parameter-Varying Systems {LPVS} 2021},
author = {Mejari, M. and Mavkov, B. and Forgione, M. and Piga, D.},
year = {2021},
doi = {},
url = {}
}
DownloadMellace, S., Kanjirangat, V., Antonucci, A. (2021). Relation clustering in narrative knowledge graphs. In AI4Narratives Workshop at 29th International Joint Conference on Artificial Intelligence and the 17th Pacific Rim International Conference on Artificial Intelligence (IJCAI-PRICAI 20).
Relation clustering in narrative knowledge graphs
Authors: Mellace, S. and Kanjirangat, V. and Antonucci, A.
Year: 2021
Abstract: When coping with literary texts such as novels or short stories, the extraction of structured information in the form of a knowledge graph might be hindered by the huge number of possible relations between the entities corresponding to the characters in the novel and the consequent hurdles in gathering supervised information about them. Such issue is addressed here as an unsupervised task empowered by transformers: relational sentences in the original text are embedded (with SBERT) and clustered in order to merge together semantically similar relations. All the sentences in the same cluster are finally summarized (with BART) and a descriptive label extracted from the summary. Preliminary tests show that such clustering might successfully detect similar relations, and provide a valuable preprocessing for semi-supervised approaches.
Published in AI4Narratives Workshop at 29th International Joint Conference on Artificial Intelligence and the 17th Pacific Rim International Conference on Artificial Intelligence (IJCAI-PRICAI 20).
Relation clustering in narrative knowledge graphs
@INPROCEEDINGS{vani2021a,
title = {Relation clustering in narrative knowledge graphs},
booktitle = {{AI4Narratives} Workshop at 29th International Joint Conference on Artificial Intelligence and the 17th Pacific Rim International Conference on Artificial Intelligence ({IJCAI}-{PRICAI} 20)},
author = {Mellace, S. and Kanjirangat, V. and Antonucci, A.},
year = {2021},
doi = {},
url = {http://ceur-ws.org/Vol-2794/paper5.pdf}
}
DownloadMichelucci, U., Sperti, M., Piga, D., Venturini, F., Deriu, M.A. (2021). A model-agnostic algorithm for bayes error determination in binary classification. Algorithms 14(11).
A model-agnostic algorithm for bayes error determination in binary classification
Authors: Michelucci, U. and Sperti, M. and Piga, D. and Venturini, F. and Deriu, M.A.
Year: 2021
Abstract: This paper presents the intrinsic limit determination algorithm (ILD Algorithm), a novel technique to determine the best possible performance, measured in terms of the AUC (area under the ROC curve) and accuracy, that can be obtained from a specific dataset in a binary classification problem with categorical features regardless of the model used. This limit, namely, the Bayes error, is completely independent of any model used and describes an intrinsic property of the dataset. The ILD algorithm thus provides important information regarding the prediction limits of any binary classification algorithm when applied to the considered dataset. In this paper, the algorithm is described in detail, its entire mathematical framework is presented and the pseudocode is given to facilitate its implementation. Finally, an example with a real dataset is given.
Published in Algorithms 14(11).
A model-agnostic algorithm for bayes error determination in binary classification
@ARTICLE{piga2021d,
title = {A model-agnostic algorithm for bayes error determination in binary classification},
journal = {Algorithms},
volume = {14},
author = {Michelucci, U. and Sperti, M. and Piga, D. and Venturini, F. and Deriu, M.A.},
number = {11},
year = {2021},
doi = {10.3390/a14110301},
url = {https://www.mdpi.com/1999-4893/14/11/301}
}
DownloadPedrero-Martin, Y., Falla, D., Martinez-Calderon, J., Liew, B.X.W., Scutari, M., Luque-Suarez, A. (2021). Self-efficacy beliefs mediate the association between pain intensity and pain interference in acute/subacute whiplash-associated disorders. European Spine Journal 20, pp. 1689–1698.
Self-efficacy beliefs mediate the association between pain intensity and pain interference in acute/subacute whiplash-associated disorders
Authors: Pedrero-Martin, Y. and Falla, D. and Martinez-Calderon, J. and Liew, B.X.W. and Scutari, M. and Luque-Suarez, A.
Year: 2021
Published in European Spine Journal 20, pp. 1689–1698.
Self-efficacy beliefs mediate the association between pain intensity and pain interference in acute/subacute whiplash-associated disorders
@ARTICLE{scutari21a,
title = {Self-efficacy beliefs mediate the association between pain intensity and pain interference in acute/subacute whiplash-associated disorders},
journal = {European Spine Journal},
volume = {20},
author = {Pedrero-Martin, Y. and Falla, D. and Martinez-Calderon, J. and Liew, B.X.W. and Scutari, M. and Luque-Suarez, A.},
pages = {1689--1698},
year = {2021},
doi = {},
url = {}
}
DownloadPiga, D, Forgione, M , Mejari, M (2021). Deep learning with transfer functions: New applications in system identification. In Proceedings of the 19th IFAC Symposium System Identification: learning models for decision and control.
Deep learning with transfer functions: New applications in system identification
Authors: Piga, D and Forgione, M and Mejari, M
Year: 2021
Abstract: This paper presents a linear dynamical operator described in terms of a rational transfer function, endowed with a well-defined and efficient back-propagation behavior for automatic derivatives computation.
The operator enables end-to-end training of structured networks containing linear transfer functions and other differentiable units by exploiting standard deep learning software.
Two relevant applications of the operator in system identification are presented. The first one consists in the integration of prediction error methods in deep learning. The dynamical operator is included as the last layer of a neural network in order to obtain the optimal one-step-ahead prediction error.
The second one considers identification of general block-oriented models from quantized data. These block-oriented models are constructed by combining linear dynamical operators with static nonlinearities described as standard feed-forward neural networks. A custom loss function corresponding to the log-likelihood of quantized output observations is defined. For gradient-based optimization, the derivatives of the log-likelihood are computed by applying the back-propagation algorithm through the whole network. Two system identification benchmarks are used to show the effectiveness of the proposed methodologies.
Published in Proceedings of the 19th IFAC Symposium System Identification: learning models for decision and control.
Deep learning with transfer functions: New applications in system identification
@INPROCEEDINGS{forgione2021c,
title = {Deep learning with transfer functions: {N}ew applications in system identification},
booktitle = {Proceedings of the 19th {IFAC} Symposium System Identification: {l}earning {m}odels for {d}ecision and {c}ontrol},
author = {Piga, D and Forgione, M and Mejari, M},
year = {2021},
doi = {},
url = {}
}
DownloadRopero, R.F., Flores, M.J., Cabañas, R., Rumí, R. (2021). A comparison bewteen elvira software and amidst toolbox in environmental data: a case study of flooding risk management. In Proceedings of the 15th Uai Conference on Bayesian Modeling Applications Workshop.
A comparison bewteen elvira software and amidst toolbox in environmental data: a case study of flooding risk management
Authors: Ropero, R.F. and Flores, M.J. and Cabañas, R. and Rumí, R.
Year: 2021
Published in Proceedings of the 15th Uai Conference on Bayesian Modeling Applications Workshop.
A comparison bewteen elvira software and amidst toolbox in environmental data: a case study of flooding risk management
@INPROCEEDINGS{cabanas2021d,
title = {A comparison bewteen elvira software and amidst toolbox in environmental data: a case study of flooding risk management},
booktitle = {Proceedings of the 15th Uai Conference on Bayesian Modeling Applications Workshop},
author = {Ropero, R.F. and Flores, M.J. and Caba\~nas, R. and Rum\'i, R.},
year = {2021},
doi = {},
url = {}
}
DownloadRoveda, L., Maggioni, B., Marescotti, E., Shahid, A., Zanchettin, M., Bemporad, A., Piga, D. (2021). Pairwise preferences-based optimization of a path-based velocity planner in robotic sealing tasks. IEEE Robotics and Automation Letters.
Pairwise preferences-based optimization of a path-based velocity planner in robotic sealing tasks
Authors: Roveda, L. and Maggioni, B. and Marescotti, E. and Shahid, A. and Zanchettin, M. and Bemporad, A. and Piga, D.
Year: 2021
Abstract: Production plants are being re-designed to im-
plement human-centered solutions. Especially considering high
added-value operations, robots are required to optimize their
behavior to achieve a task quality at least comparable to the
one obtained by the skilled operators. A manual programming
and tuning of the manipulator is not an efficient solution,
requiring to adopt towards automated strategies. Adding external
sensors (e.g., cameras) increases the robotic cell complexity and
it doesn’t solve the issue since it is usually difficult to build
explicit reward functions measuring the robot performance, while
it is easier for the user to define a qualitative comparison
between two experiments. According to these needs, in this paper,
the recently-developed preferences-based optimization approach
GLISp is employed and adapted to tune the novel developed
path-based velocity planner. The implemented solution defines
an intuitive human-centered procedure, capable of transferring
(through pairwise preferences between experiments) the task
knowledge from the operator to the manipulator. A Franka
EMIKA panda robot has been employed as a test platform to
perform a robotic sealing task (i.e., material deposition task),
validating the proposed methodology. The proposed approach
has been compared with a programming by demonstration
approach, and with the manual tuning of the path-based velocity
planner. Achieved results demonstrate the improved deposition
quality obtained with the proposed optimized path-based velocity
planner methodology in a limited number of experimental trials
(20).
Published in IEEE Robotics and Automation Letters.
Pairwise preferences-based optimization of a path-based velocity planner in robotic sealing tasks
@ARTICLE{Roveda2021c,
title = {Pairwise preferences-based optimization of a path-based velocity planner in robotic sealing tasks},
journal = {{IEEE} Robotics and Automation Letters},
author = {Roveda, L. and Maggioni, B. and Marescotti, E. and Shahid, A. and Zanchettin, M. and Bemporad, A. and Piga, D.},
year = {2021},
doi = {10.1109/LRA.2021.3094479},
url = {}
}
DownloadRoveda, L., Maroni, M., Mazzuchelli, L., Praolini, L., Bucca, G., Piga, D. (2021). Enhancing object detection performance through sensor pose Definition with bayesian optimization. In .
Enhancing object detection performance through sensor pose Definition with bayesian optimization
Authors: Roveda, L. and Maroni, M. and Mazzuchelli, L. and Praolini, L. and Bucca, G. and Piga, D.
Year: 2021
Abstract: Robots equipped with vision systems at the end-
effector provide a powerful combination in industrial contexts.
While much attention is dedicated to machine vision algorithms,
the optimization of the vision system pose is not properly
addressed (to increase object detection performance). A complete
pipeline for such optimization is proposed. To this aim, Bayesian
Optimization is employed. A Franka EMIKA Panda robot has
been used as a robotic platform, equipped at its end-effector
with an Intel© RealSense D400. Achieved results show the
high-fidelity reconstruction of the real working environment for
the offline optimization (i.e., performed simulations), together
with capabilities of the proposed Bayesian Optimization-based
approach to defining the sensor pose in a limited number of
experimental trials (50 maximum iterations has been considered).
Published in Metrology2021.
Enhancing object detection performance through sensor pose Definition with bayesian optimization
@INPROCEEDINGS{Roveda2021h,
title = {Enhancing object detection performance through sensor pose {D}efinition with bayesian optimization},
journal = {Metrology2021},
author = {Roveda, L. and Maroni, M. and Mazzuchelli, L. and Praolini, L. and Bucca, G. and Piga, D.},
year = {2021},
doi = {10.1109/METROIND4.0IOT51437.2021.9488517},
url = {}
}
DownloadRoveda, L., Piga, D. (2021). Sensorless environment stiffness and interaction force estimation for impedance control tuning in robotized interaction tasks. Autonomous Robots.
Sensorless environment stiffness and interaction force estimation for impedance control tuning in robotized interaction tasks
Authors: Roveda, L. and Piga, D.
Year: 2021
Abstract: Industrial robots are increasingly used to perform tasks requiring an interaction with the surrounding environment (e.g., assembly tasks). Such environments are usually (partially) unknown to the robot, requiring the implemented controllers to suitably react to the established interaction. Standard controllers require force/torque measurements to close the loop. However, most of the industrial manipulators do not have embedded force/torque sensor(s) and such integration results in additional costs and implementation effort. To extend the use of compliant controllers to sensorless interaction control, a model-based methodology is presented in this paper. Relying on sensorless Cartesian impedance control, two Extended Kalman Filters (EKF) are proposed: an EKF for interaction force estimation and an EKF for environment stiffness estimation. Exploiting such estimations, a control architecture is proposed to implement a sensorless force loop (exploiting the provided estimated force) with adaptive Cartesian impedance control and coupling dynamics compensation (exploiting the provided estimated environment stiffness). The described approach has been validated in both simulations and experiments. A Franka EMIKA panda robot has been used. A probing task involving different materials (i.e., with different - unknown - stiffness properties) has been considered to show the capabilities of the developed EKFs (able to converge with limited errors) and control tuning (preserving stability). Additionally, a polishing-like task and an assembly task have been implemented to show the achieved performance of the proposed methodology.
Published in Autonomous Robots.
Sensorless environment stiffness and interaction force estimation for impedance control tuning in robotized interaction tasks
@ARTICLE{Roveda2021a,
title = {Sensorless environment stiffness and interaction force estimation for impedance control tuning in robotized interaction tasks},
journal = {Autonomous Robots},
author = {Roveda, L. and Piga, D.},
year = {2021},
doi = {https://doi.org/10.1007/s10514-021-09970-z},
url = {}
}
DownloadRoveda, L., Riva, D., Bucca, G., Piga, D. (2021). External joint torques estimation for a position-controlled manipulator employing an extended kalman filter. In .
External joint torques estimation for a position-controlled manipulator employing an extended kalman filter
Authors: Roveda, L. and Riva, D. and Bucca, G. and Piga, D.
Year: 2021
Published in UbiquitousRobots2021.
External joint torques estimation for a position-controlled manipulator employing an extended kalman filter
@INPROCEEDINGS{Roveda2021i,
title = {External joint torques estimation for a position-controlled manipulator employing an extended kalman filter},
journal = {{UbiquitousRobots2021}},
author = {Roveda, L. and Riva, D. and Bucca, G. and Piga, D.},
year = {2021},
doi = {10.1109/UR52253.2021.9494674},
url = {}
}
DownloadRoveda, L., Riva, D., Bucca, G., Piga, D. (2021). Sensorless optimal switching Impact/Force controller. IEEE Access.
Sensorless optimal switching Impact/Force controller
Authors: Roveda, L. and Riva, D. and Bucca, G. and Piga, D.
Year: 2021
Abstract: Intelligent interaction control is required in many fields of application, in which different operative situations
have to be faced with different controllers. Being able to switch between optimized controllers is, indeed, of
extreme importance to maximize the task performance in the different operative conditions (i.e., free-space
motion and contact), especially when considering sensorless robots. To deal with the proposed context,
a sensorless optimal switching impact/force (OSIF) controller is proposed. The low-level robot control
is composed of an inner joint position controller, fed by an outer Cartesian impedance controller with a
reference position. The estimation of the external wrench is implemented by means of an Extended Kalman
Filter (EKF). The high-level controller (feeding the Cartesian impedance controller with the setpoint) is
composed of an optimized impact controller (LQR-based controller), an optimized force controller (SDRE-
based controller), and a continuous switching mechanism (Fuzzy Logic-based). In addition, the output of
the switching mechanism is used to adapt the Cartesian impedance control parameters (i.e., stiffness and
damping parameters). Experimental tests have been performed on a Franka EMIKA panda robot to validate
the proposed controller. Obtained results show the capabilities of the OSIF controller, being able to detect
task phase transitions while satisfying the target performance.
Published in IEEE Access.
Sensorless optimal switching Impact/Force controller
@ARTICLE{Roveda2021f,
title = {Sensorless optimal switching {Impact/Force} controller},
journal = {{IEEE} Access},
author = {Roveda, L. and Riva, D. and Bucca, G. and Piga, D.},
year = {2021},
doi = {10.1109/ACCESS.2021.3131390},
url = {}
}
DownloadRoveda, L., Shahid, A., Iannacci, N., Piga, D. (2021). Sensorless optimal interaction control exploiting environment stiffness estimation. IEEE Transactions on Control System Technology.
Sensorless optimal interaction control exploiting environment stiffness estimation
Authors: Roveda, L. and Shahid, A. and Iannacci, N. and Piga, D.
Year: 2021
Abstract: Industrial robots are increasingly used to perform tasks that require an interaction with the surrounding environment (e.g., assembly tasks). Such environments are usually (partially) unknown to the robot (in terms of dynamic characteristics), demanding the implemented controllers to suitably react to the established interaction. Standard controllers require force/torque measurements to close the loop, making it, if possible, to adapt the robot behavior to the specific environment. However, most of the industrial manipulators do not have embedded force/torque sensor(s), which entails additional effort in terms of costs and implementation for their integration in the robotic setup. To extend the use of sensorless compliant controllers to force control, a robot-environment interaction dynamics model-based methodology is presented in this article. Relying on the sensorless Cartesian impedance control, an extended Kalman filter (EKF) is proposed to estimate the stiffness of an interaction environment. Exploiting the provided estimation, the robot–environment coupled dynamic modeling is used to design an optimal LQR interaction controller to close the force loop. The control gains can be analytically computed by solving the related Riccati equation, as such gains are a function of the impedance control and environment parameters. In addition, the interaction force can be estimated to close the force loop, making the sensorless robot able to perform the target interaction task. The described approach has been validated with experiments by analyzing two scenarios: a probing task and a closing of a plastic (i.e., compliant) box with a snap-fit closure mechanism. The performance of the proposed control framework has been evaluated, highlighting the capabilities of the EKF and the optimal LQR interaction controller. Finally, the proposed control schema is enhanced by the adaptation of the EKF for the estimation of the external wrench. Two additional experiments are provided to show the improvements on the control schema (a polishing-like task and an assembly task). The Franka EMIKA panda robot has been used as the reference robotic platform for experimental validation.
Published in IEEE Transactions on Control System Technology.
Sensorless optimal interaction control exploiting environment stiffness estimation
@ARTICLE{Roveda2021b,
title = {Sensorless optimal interaction control exploiting environment stiffness estimation},
journal = {{IEEE} Transactions on Control System Technology},
author = {Roveda, L. and Shahid, A. and Iannacci, N. and Piga, D.},
year = {2021},
doi = {10.1109/TCST.2021.3061091},
url = {}
}
DownloadSelvi, D., Piga, D., Battistelli, G., Bemporad, A. (2021). Optimal direct data-driven control with stability guarantees. European Journal of Control 59.
Optimal direct data-driven control with stability guarantees
Authors: Selvi, D. and Piga, D. and Battistelli, G. and Bemporad, A.
Year: 2021
Abstract: For model-free optimal control design, this paper proposes an approach based on optimizing the reference model that is used in direct data-driven controller synthesis. Optimality is defined with respect to suitable cost functions reflecting desired performance and control objectives. We rely on the well-known Virtual Reference Feedback Tuning technique and on a direct control design approach that ensures stability of the resulting closed-loop system. The proposed design method leads to a non-convex optimization problem with a small number of variables that can be easily solved by a global optimizer, such as by particle swarm optimization. The effectiveness of the proposed solution is illustrated in simulation examples.
Published in European Journal of Control 59.
Optimal direct data-driven control with stability guarantees
@ARTICLE{piga2021b,
title = {Optimal direct data-driven control with stability guarantees},
journal = {European Journal of Control},
volume = {59},
author = {Selvi, D. and Piga, D. and Battistelli, G. and Bemporad, A.},
year = {2021},
doi = {https://doi.org/10.1016/j.ejcon.2020.09.005},
url = {}
}
DownloadShahid, A., Sesin, J., Pecioski, D., Braghin, F., Piga, D., Roveda, L. (2021). Decentralized multi-agent control of a manipulator in continuous task learning. MDPI Applied Science.
Decentralized multi-agent control of a manipulator in continuous task learning
Authors: Shahid, A. and Sesin, J. and Pecioski, D. and Braghin, F. and Piga, D. and Roveda, L.
Year: 2021
Abstract: Many real-world tasks require multiple agents to work together. When talking about
multiple agents in robotics, it is usually referenced to multiple manipulators in collaboration to
solve a given task, where each one is controlled by a single agent. However, due to the increasing
development of modular and re-configurable robots, it is also important to investigate the possibility
of implementing multi-agent controllers that learn how to manage the manipulator’s degrees of
freedom (DoF) in separated clusters for the execution of a given application (e.g., being able to
face faults or, partially, new kinematics configurations). Within this context, this paper focuses on
the decentralization of the robot control action learning and (re)execution considering a generic
multi-DoF manipulator. Indeed, the proposed framework employs a multi-agent paradigm and
investigates how such a framework impacts the control action learning process. Multiple variations of
the multi-agent framework have been proposed and tested in this research, comparing the achieved
performance w.r.t. a centralized (i.e., single-agent) control action learning framework, previously
proposed by some of the authors. As a case study, a manipulation task (i.e., grasping and lifting)
of an unknown object (to the robot controller) has been considered for validation, employing a
Franka EMIKA panda robot. The MuJoCo environment has been employed to implement and test
the proposed multi-agent framework. The achieved results show that the proposed decentralized
approach is capable of accelerating the learning process at the beginning with respect to the single-
agent framework while also reducing the computational effort. In fact, when decentralizing the
controller, it is shown that the number of variables involved in the action space can be efficiently
separated into several groups and several agents. This simplifies the original complex problem into
multiple ones, efficiently improving the task learning process.
Published in MDPI Applied Science.
Decentralized multi-agent control of a manipulator in continuous task learning
@ARTICLE{Roveda2021d,
title = {Decentralized multi-agent control of a manipulator in continuous task learning},
journal = {{MDPI} Applied Science},
author = {Shahid, A. and Sesin, J. and Pecioski, D. and Braghin, F. and Piga, D. and Roveda, L.},
year = {2021},
doi = {https://doi.org/10.3390/app112110227},
url = {}
}
DownloadSommer, M., Rüegsegger, M., Szehr, O., Del Rio, G. (2021). Deep self-optimizing artificial intelligence for tactical analysis, training and optimization . In , NATO.
Deep self-optimizing artificial intelligence for tactical analysis, training and optimization
Authors: Sommer, M. and Rüegsegger, M. and Szehr, O. and Del Rio, G.
Year: 2021
Abstract: The increasing complexity of modern multi-domain conflicts has made their tactical and strategic understanding and the identification of appropriate courses of action challenging endeavours. Modelling and simulation as part of concept development and experimentation (CD&E) provide new insight at higher speed and lower cost than what physical manoeuvres can achieve. Amongst other, human-machine teaming through computer games is a powerful means of simulating defence scenarios at various abstraction levels. However, conventional human-machine interaction is time-consuming and restricted to pre-designed scenarios, e.g., in terms of pre-programmed conditional computer actions. If one side of the game could be taken by Artificial Intelligence (AI), this would increase the diversity of explored courses of actions and lead to a more robust and comprehensive analysis. If the AI plays both sides, this allows employing the Data Farming methodology and creating and analysing a database of a large number of investigated scenarios. To achieve a high degree of variability and generalization capability of the investigated scenarios, we employ combined Reinforcement Learning and search algorithms, which have demonstrated super-human performance in various complex planning problems. Such AI systems avoid the reliance on training data, human experience and predictions by learning tactics and strategy through exploration and self-optimization. In this contribution, we present the benefits and challenges of applying a Neural-Network-based Monte Carlo Tree Search algorithm for strategic planning and training in air-defence scenarios and virtual war-gaming with systems that are available currently or potentially in the future to the Swiss Armed Forces.
Published in MSG-Symposium, NATO.
Deep self-optimizing artificial intelligence for tactical analysis, training and optimization
@INPROCEEDINGS{szehr2021b,
title = {Deep self-optimizing artificial intelligence for tactical analysis, training and optimization },
journal = {{MSG}-Symposium},
publisher = {NATO},
author = {Sommer, M. and R\"uegsegger, M. and Szehr, O. and Del Rio, G.},
year = {2021},
doi = {},
url = {}
}
DownloadSzehr, O., Zarouf, R. (2021). Explicit counterexamples to Schäffer's conjecture. Journal de Mathématiques Pures et Appliquées 146, pp. 1-30.
Explicit counterexamples to Schäffer's conjecture
Authors: Szehr, O. and Zarouf, R.
Year: 2021
Abstract: In this article we propose a new and entirely constructive approach to Schäffer's conjecture.
For full abstract please read the preprint.
Published in Journal de Mathématiques Pures et Appliquées 146, pp. 1-30.
Explicit counterexamples to Schäffer's conjecture
@ARTICLE{szehr2021a,
title = {Explicit counterexamples to {S}ch\"affer's conjecture},
journal = {Journal {d}e Math\'ematiques Pures {e}t Appliqu\'ees},
volume = {146},
author = {Szehr, O. and Zarouf, R.},
pages = {1-30},
year = {2021},
doi = {10.1016/j.matpur.2020.10.006},
url = {}
}
DownloadTermine, A., Antonucci, A., Primiero, G., Facchini, A. (2021). Logic and model checking by imprecise probabilistic interpreted systems. In Rosenfeld, A., Talmon, N. (Eds), Multi-Agent Systems. EUMAS 2021. Lecture Notes in Computer Science, Cham.
Logic and model checking by imprecise probabilistic interpreted systems
Authors: Termine, A. and Antonucci, A. and Primiero, G. and Facchini, A.
Year: 2021
Abstract: Stochastic multi-agent systems raise the necessity to extend probabilistic model checking to the epistemic domain. Results in this direction have been achieved by epistemic extensions of Probabilistic Computation Tree Logic and related Probabilistic Interpreted Systems. The latter, however, suffer of an important limitation: they require the probabilities governing the system’s behaviour to be fully specified. A promising way to overcome this limitation is represented by imprecise probabilities. In this paper we introduce imprecise probabilistic inter- preted systems and present a related logical language and model-checking procedures based on recent advances in the study of imprecise Markov processes.
Published in Rosenfeld, A., Talmon, N. (Eds), Multi-Agent Systems. EUMAS 2021. Lecture Notes in Computer Science, Cham.
Logic and model checking by imprecise probabilistic interpreted systems
@INPROCEEDINGS{antonucci2021d,
title = {Logic and model checking by imprecise probabilistic interpreted systems},
editor = {Rosenfeld, A. and Talmon, N.},
address = {Cham},
booktitle = {Multi-Agent Systems. {EUMAS} 2021. Lecture Notes in Computer Science},
author = {Termine, A. and Antonucci, A. and Primiero, G. and Facchini, A.},
year = {2021},
doi = {https://doi.org/10.1007/978-3-030-82254-5_13},
url = {}
}
DownloadTermine, A., Antonucci, A., Facchini, A., Primiero, G. (2021). Robust model checking with imprecise markov reward models. In De Bock, J., Cano, A., Miranda, E., Moral, S. (Ed), ISIPTA 2021, PMLR 147, JMLR.org, pp. 299–309.
Robust model checking with imprecise markov reward models
Authors: Termine, A. and Antonucci, A. and Facchini, A. and Primiero, G.
Year: 2021
Abstract: In recent years probabilistic model checking has be- come an important area of research because of the diffusion of computational systems of stochastic nature. Despite its great success, standard probabilistic model checking suffers the limitation of requiring a sharp specification of the probabilities governing the model behaviour. The theory of imprecise probabilities offers a natural approach to overcome such limitation by a sensitivity analysis with respect to the values of these parameters. However, only extensions based on discrete-time imprecise Markov chains have been con- sidered so far for such a robust approach to model checking. We present a further extension based on imprecise Markov reward models. In particular, we derive efficient algorithms to compute lower and upper bounds of the expected cumulative reward and proba- bilistic bounded rewards based on existing results for imprecise Markov chains. These ideas are tested on a real case study involving the spend-down costs of geriatric medicine departments.
Published in De Bock, J., Cano, A., Miranda, E., Moral, S. (Ed), ISIPTA 2021, PMLR 147, JMLR.org, pp. 299–309.
Robust model checking with imprecise markov reward models
@INPROCEEDINGS{antonucci2021a,
title = {Robust model checking with imprecise markov reward models},
editor = {De Bock, J., Cano, A., Miranda, E., Moral, S.},
publisher = {JMLR.org},
series = {PMLR},
volume = {147},
booktitle = {{ISIPTA} 2021},
author = {Termine, A. and Antonucci, A. and Facchini, A. and Primiero, G.},
pages = {299--309},
year = {2021},
doi = {},
url = {https://leo.ugr.es/isipta21/pmlr/termine21.pdf}
}
DownloadZaffalon, M., Antonucci, A., Cabañas, R. (2021). Causal expectation-maximisation. In WHY-21 @ NeurIPS 2021.
Causal expectation-maximisation
Authors: Zaffalon, M. and Antonucci, A. and Cabañas, R.
Year: 2021
Abstract: Structural causal models are the basic modelling unit in Pearl's causal theory; in principle they allow us to solve counterfactuals, which are at the top rung of the ladder of causation. But they often contain latent variables that limit their application to special settings. This appears to be a consequence of the fact, proven in this paper, that causal inference is NP-hard even in models characterised by polytree-shaped graphs. To deal with such a hardness, we introduce the causal EM algorithm. Its primary aim is to reconstruct the uncertainty about the latent variables from data about categorical manifest variables. Counterfactual inference is then addressed via standard algorithms for Bayesian networks. The result is a general method to approximately compute counterfactuals, be they identifiable or not (in which case we deliver bounds). We show empirically, as well as by deriving credible intervals, that the approximation we provide becomes accurate in a fair number of EM runs. These results lead us finally to argue that there appears to be an unnoticed limitation to the trending idea that counterfactual bounds can often be computed without knowledge of the structural equations.
Published in WHY-21 @ NeurIPS 2021.
Causal expectation-maximisation
@INPROCEEDINGS{zaffalon2021c,
title = {Causal expectation-maximisation},
booktitle = {{WHY}-21 @ {NeurIPS} 2021},
author = {Zaffalon, M. and Antonucci, A. and Cabañas, R.},
year = {2021},
doi = {},
url = {https://why21.causalai.net/papers/WHY21_52.pdf}
}
DownloadZaffalon, M., Miranda, E. (2021). Desirability foundations of robust rational decision making. Synthese 198(27), pp. S6529–S6570.
Desirability foundations of robust rational decision making
Authors: Zaffalon, M. and Miranda, E.
Year: 2021
Abstract: Recent work has formally linked the traditional axiomatisation of incomplete preferences à la Anscombe-Aumann with the theory of desirability developed in the context of imprecise probability, by showing in particular that they are the very same theory. The equivalence has been established under the constraint that the set of possible prizes is finite. In this paper, we relax such a constraint, thus de facto creating one of the most general theories of rationality and decision making available today. We provide the theory with a sound interpretation and with basic notions, and results, for the separation of beliefs and values, and for the case of complete preferences. Moreover, we discuss the role of conglomerability for the presented theory, arguing that it should be a rationality requirement under very broad conditions.
Published in Synthese 198(27), Springer, pp. S6529–S6570.
Note: published online in 2018.
Desirability foundations of robust rational decision making
@ARTICLE{zaffalon2019a,
title = {Desirability foundations of robust rational decision making},
journal = {Synthese},
publisher = {Springer},
volume = {198},
author = {Zaffalon, M. and Miranda, E.},
number = {27},
pages = {S6529--S6570},
year = {2021},
doi = {10.1007/s11229-018-02010-x},
url = {}
}
DownloadZaffalon, M., Miranda, E. (2021). The sure thing. In De Bock, J., Cano, A., Miranda, E., Moral, S. (Ed), ISIPTA 2021, PMLR 147, JMLR.org, pp. 342–351.
The sure thing
Authors: Zaffalon, M., Miranda, E.
Year: 2021
Abstract: If we prefer action a to b both under an event and under its complement, then we should just prefer a to b. This is Savage's sure-thing principle. In spite of its intuitive- and simple-looking nature, for which it gets almost immediate acceptance, the sure thing is not a logical principle. So where does it get its support from? In fact, the sure thing may actually fail. This is related to a variety of deep and foundational concepts in causality, decision theory, and probability, as well as to Simpsons’ paradox and Blyth’s game. In this paper we try to systematically clarify such a network of relations. Then we propose a general desirability theory for nonlinear utility scales. We use that to show that the sure thing is primitive to many of the previous concepts: In non-causal settings, the sure thing follows from considerations of temporal coherence and coincides with conglomerability; it can be understood as a rationality axiom to enable well-behaved conditioning in logic. In causal settings, it can be derived using only coherence and a causal independence condition.
Accepted in De Bock, J., Cano, A., Miranda, E., Moral, S. (Ed), ISIPTA 2021, PMLR 147, JMLR.org, pp. 342–351.
The sure thing
@INPROCEEDINGS{zaffalon2021a,
title = {The sure thing},
editor = {De Bock, J., Cano, A., Miranda, E., Moral, S.},
publisher = {JMLR.org},
series = {PMLR},
volume = {147},
booktitle = {{ISIPTA} 2021},
author = {Zaffalon, M., Miranda, E.},
pages = {342--351},
year = {2021},
doi = {},
url = {https://www.sipta.org/isipta21/pmlr/zaffalon21.pdf}
}
DownloadZhu, M., Bemporad, A., Piga, D. (2021). Preference-based MPC calibration. In Proceedings of the European Control Conference, Napoli, Italy.
Preference-based MPC calibration
Authors: Zhu, M. and Bemporad, A. and Piga, D.
Year: 2021
Abstract: Automating the calibration of the parameters of a control policy by means of global optimization requires quantifying a closed-loop performance function. As this can be impractical in many situations, in this paper we suggest a semiautomated calibration approach that requires instead a human calibrator to express a preference on whether a certain control policy is “better” than another one, therefore eliminating the need of an explicit performance index. In particular, we focus our attention on semi-automated calibration of Model Predictive Controllers (MPCs), for which we attempt computing the set of best calibration parameters by employing the recently-developed active preference-based optimization algorithm GLISp. Based on the preferences expressed by the human operator, GLISp learns a surrogate of the underlying closed-loop performance index that the calibrator (unconsciously) uses and proposes, iteratively, a new set of calibration parameters to him or her for testing and for comparison against previous experimental results. The resulting semi-automated calibration procedure is tested on two case studies, showing the capabilities of the approach in achieving near-optimal performance within a limited number of experiments.
Published in Proceedings of the European Control Conference, Napoli, Italy.
Preference-based MPC calibration
@INPROCEEDINGS{piga2021e,
title = {Preference-based {MPC} calibration},
address = {Napoli, Italy},
booktitle = {Proceedings of the European Control Conference},
author = {Zhu, M. and Bemporad, A. and Piga, D.},
year = {2021},
doi = {},
url = {}
}
Download top2020
Akata, Z., Ballier, D., de Rijke, M., Dignum, F., Dignum, V., Eiben, G., Fokkens, A., Grossi, D., Hindriks, K., Hoos, H., Hung, H., Jonker, C., Monz, C., Neerincx, M., Oliehoek, F., Prakken, H., Schlobach, S., van der Gaag, L.C., van Harmelen, F., van Hoof, H., van Riemsdijk, B., van Wynsberghe, A., Verbrugge, R., Berheij, B., Vossen, P., Welling, M. (2020). A research agenda for hybrid intelligence: Augmenting human intellect with collaborative, adaptive, responsible and explainable artificial intelligence. IEEE Computer 53, pp. 18-28.
A research agenda for hybrid intelligence: Augmenting human intellect with collaborative, adaptive, responsible and explainable artificial intelligence
Authors: Akata, Z. and Ballier, D. and de Rijke, M. and Dignum, F. and Dignum, V. and Eiben, G. and Fokkens, A. and Grossi, D. and Hindriks, K. and Hoos, H. and Hung, H. and Jonker, C. and Monz, C. and Neerincx, M. and Oliehoek, F. and Prakken, H. and Schlobach, S. and van der Gaag, L.C. and van Harmelen, F. and van Hoof, H. and van Riemsdijk, B. and van Wynsberghe, A. and Verbrugge, R. and Berheij, B. and Vossen, P. and Welling, M.
Year: 2020
Abstract: We define Hybrid Intelligence (HI) as the combination of human and machine intelligence, augmenting human intellect and capabilities instead of replacing them, and achieve goals that were unreachable by either humans or machines alone. We identify Hybrid Intelligence as an important new research focus for the field of Artificial Intelligence, and we set a research agenda for Hybrid Intelligence by formulating four challenges: how do we develop AI systems that work in synergy with humans (Collaborative HI); how can these systems learn from and adapt to humans and their environment (Adaptive HI); how do we ensure that they behave ethically and responsibly (Responsible HI); and how can AI systems and humans share and explain their awareness, goals and strategies (Explainable HI). For each of these four challenges, we survey the state of the art and define a research agenda.
Published in IEEE Computer 53, pp. 18-28.
A research agenda for hybrid intelligence: Augmenting human intellect with collaborative, adaptive, responsible and explainable artificial intelligence
@ARTICLE{Linda2020b,
title = {A research agenda for hybrid intelligence: {A}ugmenting human intellect with collaborative, adaptive, responsible and explainable artificial intelligence},
journal = {{IEEE} Computer},
volume = {53},
author = {Akata, Z. and Ballier, D. and de Rijke, M. and Dignum, F. and Dignum, V. and Eiben, G. and Fokkens, A. and Grossi, D. and Hindriks, K. and Hoos, H. and Hung, H. and Jonker, C. and Monz, C. and Neerincx, M. and Oliehoek, F. and Prakken, H. and Schlobach, S. and van der Gaag, L.C. and van Harmelen, F. and van Hoof, H. and van Riemsdijk, B. and van Wynsberghe, A. and Verbrugge, R. and Berheij, B. and Vossen, P. and Welling, M.},
pages = {18-28},
year = {2020},
doi = {10.1109/MC.2020.2996587},
url = {}
}
DownloadAntonucci, A., Tiotto, T. (2020). Approximate MMAP by marginal search. In Brawner, K.W., Barták, R., Bell, E. (Eds), Proceedings of the Thirty-third International Florida Artificial Intelligence Research Society Conference (FLAIRS-33), AAAI Press, North Miami Beach, Florida, USA, pp. 181–184.
Approximate MMAP by marginal search
Authors: Antonucci, A. and Tiotto, T.
Year: 2020
Abstract: We present a heuristic strategy for marginal MAP (MMAP) queries in graphical models. The algorithm is based on a reduction of the task to a polynomial number of marginal inference computations. Given an input evidence, the marginals mass functions of the variables to be explained are computed. Marginal information gain is used to decide the variables to be explained first, and their most probable marginal states are consequently moved to the evidence. The sequential iteration of this procedure leads to a MMAP explanation and the minimum information gain obtained during the process can be regarded as a confidence measure for the explanation. Preliminary experiments show that the proposed confidence measure is properly detecting instances for which the algorithm is accurate and, for sufficiently high confidence levels, the algorithm gives the exact solution or an approximation whose Hamming distance from the exact one is small.
Published in Brawner, K.W., Barták, R., Bell, E. (Eds), Proceedings of the Thirty-third International Florida Artificial Intelligence Research Society Conference (FLAIRS-33), AAAI Press, North Miami Beach, Florida, USA, pp. 181–184.
Approximate MMAP by marginal search
@INPROCEEDINGS{antonucci2020a,
title = {Approximate {MMAP} by marginal search},
editor = {Brawner, K.W. and Bart\'ak, R. and Bell, E.},
publisher = {AAAI Press},
address = {North Miami Beach, Florida, USA},
booktitle = {Proceedings of the Thirty-{t}hird International Florida Artificial Intelligence Research Society Conference ({FLAIRS}-33)},
author = {Antonucci, A. and Tiotto, T.},
pages = {181--184},
year = {2020},
doi = {},
url = {https://www.aaai.org/Library/FLAIRS/flairs20contents.php}
}
DownloadAzzimonti, L., Corani, G., Scutari, M. (2020). Structure learning from related data sets with a hierarchical Bayesian score. In Proceedings of the 10th International Conference on Probabilistic Graphical Models (PGM 2020) 138, PMLR, pp. 5–16.
Structure learning from related data sets with a hierarchical Bayesian score
Authors: Azzimonti, L. and Corani, G. and Scutari, M.
Year: 2020
Abstract: Score functions for learning the structure of Bayesian networks in the literature assume that data are a homogeneous set of observations; whereas it is often the case that they comprise different related, but not homogeneous, data sets collected in different ways. In this paper we propose a new Bayesian Dirichlet score, which we call Bayesian Hierarchical Dirichlet (BHD). The proposed score is based on a hierarchical model that pools information across data sets to learn a single encompassing network structure, while taking into account the differences in their probabilistic structures. We derive a closed-form expression for BHD using a variational approximation of the marginal likelihood and we study its performance using simulated data. We find that, when data comprise multiple related data sets, BHD outperforms the Bayesian Dirichlet equivalent uniform (BDeu) score in terms of reconstruction accuracy as measured by the Structural Hamming distance, and that it is as accurate as BDeu when data are homogeneous. Moreover, the estimated networks are sparser and therefore more interpretable than those obtained with BDeu, thanks to a lower number of false positive arcs.
Published in Proceedings of the 10th International Conference on Probabilistic Graphical Models (PGM 2020) 138, PMLR, pp. 5–16.
Structure learning from related data sets with a hierarchical Bayesian score
@INPROCEEDINGS{azzimonti2020a,
title = {Structure learning from related data sets with a hierarchical {B}ayesian score},
publisher = {PMLR},
volume = {138},
booktitle = {Proceedings of the 10th International Conference on Probabilistic Graphical Models ({PGM} 2020)},
author = {Azzimonti, L. and Corani, G. and Scutari, M.},
pages = {5--16},
year = {2020},
doi = {},
url = {http://proceedings.mlr.press/v138/azzimonti20a.html}
}
DownloadAzzimonti, D., Rottondi, C., Giusti, A., Tornatore, M., Bianco, A. (2020). Active vs transfer learning approaches for qot estimation with small training datasets. In Optical Fiber Communication Conference (OFC 2020), Optical Society of America, M4E.1.
Active vs transfer learning approaches for qot estimation with small training datasets
Authors: Azzimonti, D. and Rottondi, C. and Giusti, A. and Tornatore, M. and Bianco, A.
Year: 2020
Abstract: We compare the level of accuracy achieved by active learning and domain adaptation approaches for quality of transmission estimation of an unestablished lightpath, in presence of small-sized training datasets.
Published in Optical Fiber Communication Conference (OFC 2020), Optical Society of America, M4E.1.
Active vs transfer learning approaches for qot estimation with small training datasets
@INPROCEEDINGS{azzimontid2020b,
title = {Active vs transfer learning approaches for qot estimation with small training datasets},
publisher = {Optical Society of America},
booktitle = {Optical Fiber Communication Conference ({OFC} 2020)},
author = {Azzimonti, D. and Rottondi, C. and Giusti, A. and Tornatore, M. and Bianco, A.},
pages = {M4E.1},
year = {2020},
doi = {10.1364/OFC.2020.M4E.1},
url = {}
}
DownloadAzzimonti, D., Rottondi, C., Tornatore, M. (2020). Reducing probes for quality of transmission estimation in optical networks with active learning. J. Opt. Commun. Netw. 12(1), pp. A38–A48.
Reducing probes for quality of transmission estimation in optical networks with active learning
Authors: Azzimonti, D. and Rottondi, C. and Tornatore, M.
Year: 2020
Abstract: Estimating the quality of transmission (QoT) of a lightpath before its establishment is a critical procedure for efficient design and management of optical networks. Recently, supervised machine learning (ML) techniques for QoT estimation have been proposed as an effective alternative to well-established, yet approximated, analytic models that often require the introduction of conservative margins to compensate for model inaccuracies and uncertainties. Unfortunately, to ensure high estimation accuracy, the training set (i.e., the set of historical field data, or “samples,” required to train these supervised ML algorithms) must be very large, while in real network deployments, the number of monitored/monitorable lightpaths is limited by several practical considerations. This is especially true for lightpaths with an above-threshold bit error rate (BER) (i.e., malfunctioning or wrongly dimensioned lightpaths), which are infrequently observed during network operation. Samples with above-threshold BERs can be acquired by deploying probe lightpaths, but at the cost of increased operational expenditures and wastage of spectral resources. In this paper, we propose to use active learning to reduce the number of probes needed for ML-based QoT estimation. We build an estimation model based on Gaussian processes, which allows iterative identification of those QoT instances that minimize estimation uncertainty. Numerical results using synthetically generated datasets show that, by using the proposed active learning approach, we can achieve the same performance of standard offline supervised ML methods, but with a remarkable reduction (at least 5% and up to 75%) in the number of training samples.
Published in J. Opt. Commun. Netw. 12(1), OSA, pp. A38–A48.
Reducing probes for quality of transmission estimation in optical networks with active learning
@ARTICLE{azzimontid2020a,
title = {Reducing probes for quality of transmission estimation in optical networks with active learning},
journal = {J. Opt. Commun. Netw.},
publisher = {OSA},
volume = {12},
author = {Azzimonti, D. and Rottondi, C. and Tornatore, M.},
number = {1},
pages = {A38--A48},
year = {2020},
doi = {10.1364/JOCN.12.000A38},
url = {}
}
DownloadBenavoli, A., Azzimonti, D., Piga, D. (2020). Skew gaussian processes for classification. Machine Learning 109.
Skew gaussian processes for classification
Authors: Benavoli, A. and Azzimonti, D. and Piga, D.
Year: 2020
Abstract: Gaussian processes (GPs) are distributions over functions, which provide a Bayesian nonparametric approach to regression and classification. In spite of their success, GPs have limited use in some applications, for example, in some cases a symmetric distribution with respect to its mean is an unreasonable model. This implies, for instance, that the mean and the median coincide, while the mean and median in an asymmetric (skewed) distribution can be different numbers. In this paper, we propose skew-Gaussian processes (SkewGPs) as a non-parametric prior over functions. A SkewGP extends the multivariate unified skew-normal distribution over finite dimensional vectors to a stochastic processes. The SkewGP class of distributions includes GPs and, therefore, SkewGPs inherit all good properties of GPs and increase their flexibility by allowing asymmetry in the probabilistic model. By exploiting the fact that SkewGP and probit likelihood are conjugate model, we derive closed form expressions for the marginal likelihood and predictive distribution of this new nonparametric classifier. We verify empirically that the proposed SkewGP classifier provides a better performance than a GP classifier based on either Laplace’s method or expectation propagation.
Published in Machine Learning 109.
Skew gaussian processes for classification
@ARTICLE{azzimontid2020c,
title = {Skew gaussian processes for classification},
journal = {Machine Learning},
volume = {109},
author = {Benavoli, A. and Azzimonti, D. and Piga, D.},
year = {2020},
doi = {10.1007/s10994-020-05906-3},
url = {}
}
DownloadBodewes, T., Scutari, M. (2020). Identifiability and consistency of bayesian network structure learning from incomplete data. Proceedings of Machine Learning Research (PGM 2020) 138, pp. 29–40.
Identifiability and consistency of bayesian network structure learning from incomplete data
Authors: Bodewes, T. and Scutari, M.
Year: 2020
Published in Proceedings of Machine Learning Research (PGM 2020) 138, pp. 29–40.
Identifiability and consistency of bayesian network structure learning from incomplete data
@ARTICLE{scutari20c,
title = {Identifiability and consistency of bayesian network structure learning from incomplete data},
journal = {Proceedings of Machine Learning Research ({PGM} 2020)},
volume = {138},
author = {Bodewes, T. and Scutari, M.},
pages = {29--40},
year = {2020},
doi = {},
url = {}
}
DownloadBregoli, A., Scutari, M., Stella, F. (2020). Constraint-based learning for continuous-time bayesian networks. Proceedings of Machine Learning Research (PGM 2020) 138, pp. 41–52.
Constraint-based learning for continuous-time bayesian networks
Authors: Bregoli, A. and Scutari, M. and Stella, F.
Year: 2020
Published in Proceedings of Machine Learning Research (PGM 2020) 138, pp. 41–52.
Note: Best Student Paper award.
Constraint-based learning for continuous-time bayesian networks
@ARTICLE{scutari20g,
title = {Constraint-based learning for continuous-time bayesian networks},
journal = {Proceedings of Machine Learning Research ({PGM} 2020)},
volume = {138},
author = {Bregoli, A. and Scutari, M. and Stella, F.},
pages = {41--52},
year = {2020},
doi = {},
url = {}
}
DownloadBriganti, G., Scutari, M., Linkowski, P. (2020). A machine learning approach to relationships among alexithymia components. Psychiatria Danubina 32(Suppl. 1), pp. 180–187.
A machine learning approach to relationships among alexithymia components
Authors: Briganti, G. and Scutari, M. and Linkowski, P.
Year: 2020
Published in Psychiatria Danubina 32(Suppl. 1), pp. 180–187.
A machine learning approach to relationships among alexithymia components
@ARTICLE{scutari20e,
title = {A machine learning approach to relationships among alexithymia components},
journal = {Psychiatria Danubina},
volume = {32},
author = {Briganti, G. and Scutari, M. and Linkowski, P.},
number = {Suppl. 1},
pages = {180--187},
year = {2020},
doi = {},
url = {}
}
DownloadCabañas, R., Antonucci, A., Huber, D., Zaffalon, M. (2020). Credici: a java library for causal inference by credal networks. In Proceedings of the 10th International Conference on Probabilistic Graphical Models, Proceedings of Machine Learning Research, PMLR, Aalborg, Denmark.
Credici: a java library for causal inference by credal networks
Authors: Cabañas, R. and Antonucci, A. and Huber, D. and Zaffalon, M.
Year: 2020
Abstract: We present CREDICI, a Java open-source tool for causal inference based on credal networks. Credal networks are an extension of Bayesian networks where local probability mass functions are only constrained to belong to given, so-called \emphcredal, sets. CREDICI is based on the recent work of Zaffalon et al. (2020), where an equivalence between Pearl's structural causal models and credal networks has been derived. This allows to reduce a counterfactual query in a causal model to a standard query in a credal network, even in the case of unidentifiable causal effects. The necessary transformations and data structures are implemented in CREDICI, while inferences are eventually computed by CREMA (Huber et al., 2020), a twin library for general credal network inference. Here we discuss the main implementation challenges and possible outlooks.
Published in Proceedings of the 10th International Conference on Probabilistic Graphical Models, Proceedings of Machine Learning Research, PMLR, Aalborg, Denmark.
Credici: a java library for causal inference by credal networks
@INPROCEEDINGS{cabanas2020a,
title = {Credici: a java library for causal inference by credal networks},
publisher = {PMLR},
address = {Aalborg, Denmark},
series = {Proceedings of Machine Learning Research},
booktitle = {Proceedings of the 10th International Conference on Probabilistic Graphical Models},
author = {Caba\~nas, R. and Antonucci, A. and Huber, D. and Zaffalon, M.},
year = {2020},
doi = {},
url = {}
}
DownloadCabañas, R., Cózar, J., Salmerón, A., Masegosa, A.R. (2020). Probabilistic graphical models with neural networks in inferpy. In Proceedings of the 10th International Conference on Probabilistic Graphical Models, Proceedings of Machine Learning Research, PMLR, Aalborg, Denmark.
Probabilistic graphical models with neural networks in inferpy
Authors: Cabañas, R. and Cózar, J. and Salmerón, A. and Masegosa, A.R.
Year: 2020
Abstract: InferPy is an open-source Python package for variational inference in probabilistic models containing neural networks. Other similar libraries are often difficult for non-expert users. InferPy provides a much more compact and simple way to code such models, at the expense of slightly reducing expressibility and flexibility. The main objective of this package is to permit its use without having a strong theoretical background or thorough knowledge of the deep learning frameworks.
Published in Proceedings of the 10th International Conference on Probabilistic Graphical Models, Proceedings of Machine Learning Research, PMLR, Aalborg, Denmark.
Probabilistic graphical models with neural networks in inferpy
@INPROCEEDINGS{cabanas2020b,
title = {Probabilistic graphical models with neural networks in inferpy},
publisher = {PMLR},
address = {Aalborg, Denmark},
series = {Proceedings of Machine Learning Research},
booktitle = {Proceedings of the 10th International Conference on Probabilistic Graphical Models},
author = {Caba\~nas, R. and C\'ozar, J. and Salmer\'on, A. and Masegosa, A.R.},
year = {2020},
doi = {},
url = {}
}
DownloadCannelli, L., Facchinei, F., Scutari, G., Kungurtsev, V. (2020). Asynchronous optimization over graphs: linear convergence under error bound conditions. IEEE Transactions on Automatic Control 66(10), pp. 4604-4619.
Asynchronous optimization over graphs: linear convergence under error bound conditions
Authors: Cannelli, L. and Facchinei, F. and Scutari, G. and Kungurtsev, V.
Year: 2020
Abstract: We consider convex and nonconvex constrained optimization with a partially separable objective function: agents minimize the sum of local objective functions, each of which is known only by the associated agent and depends on the variables of that agent and those of a few others. This partitioned setting arises in several applications of practical interest. We propose what is, to the best of our knowledge, the first distributed, asynchronous algorithm with rate guarantees for this class of problems. When the objective function is nonconvex, the algorithm provably converges to a stationary solution at a sublinear rate whereas linear rate is achieved under the renowned Luo-Tseng error bound condition (which is less stringent than strong convexity). Numerical results on matrix completion and LASSO problems show the effectiveness of our method.
Published in IEEE Transactions on Automatic Control 66(10), pp. 4604-4619.
Asynchronous optimization over graphs: linear convergence under error bound conditions
@ARTICLE{cannelli2020a,
title = {Asynchronous optimization over graphs: linear convergence under error bound conditions},
journal = {{IEEE} Transactions on Automatic Control},
volume = {66},
author = {Cannelli, L. and Facchinei, F. and Scutari, G. and Kungurtsev, V.},
number = {10},
pages = {4604-4619},
year = {2020},
doi = {10.1109/TAC.2020.3033490},
url = {}
}
DownloadCasanova, A., Miranda, E., Zaffalon, M. (2020). Social pooling of beliefs and values with desirability. Proceedings of the 33rd International Flairs Conference (FLAIRS-33).
Social pooling of beliefs and values with desirability
Authors: Casanova, A. and Miranda, E. and Zaffalon, M.
Year: 2020
Abstract: The problem of aggregating beliefs and values of rational
subjects is treated with the formalism of sets of desirable
gambles. This leads on the one hand to a new perspective
of traditional results of social choice (in particular Arrow’s
theorem as well as sufficient conditions for the existence of
an oligarchy and democracy) and on the other hand to use the
same framework to create connections with opinion pooling.
In particular, we show that weak Pareto can be derived as a
coherence requirement and discuss the aggregation of state
independent beliefs.
Introduction
Accepted in Proceedings of the 33rd International Flairs Conference (FLAIRS-33).
Social pooling of beliefs and values with desirability
@ARTICLE{casanova2020a,
title = {Social pooling of beliefs and values with desirability},
journal = {Proceedings of the 33rd International Flairs Conference ({FLAIRS}-33)},
author = {Casanova, A. and Miranda, E. and Zaffalon, M.},
year = {2020},
doi = {},
url = {https://www.flairs-33.info/}
}
DownloadColetti, G., van der Gaag, L.C., Petturiti, D., Vantaggi, B. (2020). Detecting correlation between extreme probability events. International Journal of General Systems 49(1), pp. 64–87.
Detecting correlation between extreme probability events
Authors: Coletti, G. and van der Gaag, L.C. and Petturiti, D. and Vantaggi, B.
Year: 2020
Abstract: As classical definitions of correlation give rise to counterintuitive statements when extreme probability events are involved, we introduce enhanced notions of positive and negative correlation in the general framework of coherent conditional probability. These notions allow to handle extreme probability events in a principled way by accommodating the different levels of strength of the zero probabilities involved (namely, zero layers). Since the detection of correlations by means of zero layers is computationally challenging, we provide a full characterization relying on only conditional probability values.
Keywords: Conditional probability, Extreme probability event, Coherence, Correlation
Published in International Journal of General Systems 49(1), pp. 64–87.
Detecting correlation between extreme probability events
@ARTICLE{Linda2019e,
title = {Detecting correlation between extreme probability events},
journal = {International Journal of General Systems },
volume = {49},
author = {Coletti, G. and van der Gaag, L.C. and Petturiti, D. and Vantaggi, B.},
number = {1},
pages = {64--87},
year = {2020},
doi = {10.1080/03081079.2019.1692005},
url = {}
}
DownloadCorani, G., Azzimonti, D., Augusto, J.P.S.C., Zaffalon, M. (2020). Probabilistic reconciliation of hierarchical forecast via Bayes’ rule. In Hutter, Frank, Kersting, Kristian, Lijffijt, Jefrey, Valera, Isabel (Eds), Joint European Conference on Machine Learning and Knowledge Discovery in Database (ECML- PKDD), pp. 211–226.
Probabilistic reconciliation of hierarchical forecast via Bayes’ rule
Authors: Corani, G. and Azzimonti, D. and Augusto, J.P.S.C. and Zaffalon, M.
Year: 2020
Abstract: We present a novel approach for reconciling hierarchical forecasts, based on Bayes’ rule. We define a prior distribution for the bottom time series of the hierarchy, based on the bottom base forecasts. Then we update their distribution via Bayes’ rule, based on the base forecasts
for the upper time series. Under the Gaussian assumption, we derive the updating in closed-form. We derive two algorithms, which differ as for the assumed independencies. We discuss their relation with the MinT reconciliation algorithm and with the Kalman filter, and we compare them experimentally.
Published in Hutter, Frank, Kersting, Kristian, Lijffijt, Jefrey, Valera, Isabel (Eds), Joint European Conference on Machine Learning and Knowledge Discovery in Database (ECML- PKDD), pp. 211–226.
Probabilistic reconciliation of hierarchical forecast via Bayes’ rule
@INPROCEEDINGS{corani2020a,
title = {Probabilistic reconciliation of hierarchical forecast via {B}ayes’ rule},
editor = {Hutter, Frank and Kersting, Kristian and Lijffijt, Jefrey and Valera, Isabel},
booktitle = {Joint European Conference on Machine Learning and Knowledge Discovery in Database ({ECML}- {PKDD})},
author = {Corani, G. and Azzimonti, D. and Augusto, J.P.S.C. and Zaffalon, M.},
pages = {211--226},
year = {2020},
doi = {10.1007/978-3-030-67664-3_13},
url = {}
}
DownloadCózar, J., Cabañas, R., Salmerón, A., Masegosa, A.R. (2020). Inferpy: probabilistic modeling with deep neural networks made easy. Neurocomputing 415, pp. 408–410.
Inferpy: probabilistic modeling with deep neural networks made easy
Authors: Cózar, J. and Cabañas, R. and Salmerón, A. and Masegosa, A.R.
Year: 2020
Abstract: InferPy is a Python package for probabilistic modeling with deep neural networks. It defines a user-friendly API that trades-off model complexity with ease of use, unlike other libraries whose focus is on dealing with very general probabilistic models at the cost of having a more complex API. In particular, this package allows to define, learn and evaluate general hierarchical probabilistic models containing deep neural networks in a compact and simple way. InferPy is built on top of Tensorflow Probability and Keras.
Published in Neurocomputing 415, pp. 408–410.
Inferpy: probabilistic modeling with deep neural networks made easy
@ARTICLE{cabanas2020c,
title = {Inferpy: probabilistic modeling with deep neural networks made easy},
journal = {Neurocomputing},
volume = {415},
author = {C\'ozar, J. and Caba\~nas, R. and Salmer\'on, A. and Masegosa, A.R.},
pages = {408--410},
year = {2020},
doi = {https://doi.org/10.1016/j.neucom.2020.07.117},
url = {http://www.sciencedirect.com/science/article/pii/S092523122031328X}
}
DownloadFisher, H., Gittoes, M., Evans, L., Bitchell, L., Mullen, R., Scutari, M. (2020). An interdisciplinary examination of stress and injury occurrence in athletes. Frontiers in Sports and Active Living 2(595619), pp. 1–20.
An interdisciplinary examination of stress and injury occurrence in athletes
Authors: Fisher, H. and Gittoes, M. and Evans, L. and Bitchell, L. and Mullen, R. and Scutari, M.
Year: 2020
Published in Frontiers in Sports and Active Living 2(595619), pp. 1–20.
An interdisciplinary examination of stress and injury occurrence in athletes
@ARTICLE{scutari20f,
title = {An interdisciplinary examination of stress and injury occurrence in athletes},
journal = {Frontiers in Sports and Active Living},
volume = {2},
author = {Fisher, H. and Gittoes, M. and Evans, L. and Bitchell, L. and Mullen, R. and Scutari, M.},
number = {595619},
pages = {1--20},
year = {2020},
doi = {},
url = {}
}
DownloadForgione, M., Piga, D. (2020). Model structures and fitting criteria for system identification with neural networks. In Proceedings of the 14th IEEE International Conference Application of Information and Communication Technologies (AICT 20).
Model structures and fitting criteria for system identification with neural networks
Authors: Forgione, M. and Piga, D.
Year: 2020
Abstract: This paper focuses on the identification of dynamical systems with tailor-made model structures, where neural networks are used to approximate uncertain components and domain knowledge is retained, if avail-able. These model structures are fitted to measured data using different criteria including a computationally efficient approach minimizing a regularized multi-step ahead simulation error. In this approach, the neural network parameters are estimated along with the initial conditions used to simulate the output signal in small-size subsequences. A regularization term is included in the fitting cost in order to enforce these initial conditions to be consistent with the estimated system dynamics. Pitfalls and limitations of naive one-step prediction and simulation error minimization are also discussed.
Published in Proceedings of the 14th IEEE International Conference Application of Information and Communication Technologies (AICT 20).
Model structures and fitting criteria for system identification with neural networks
@INPROCEEDINGS{forgione2020b,
title = {Model structures and fitting criteria for system identification with neural networks},
booktitle = {Proceedings of the 14th {IEEE} International Conference Application of Information and Communication Technologies ({AICT} 20)},
author = {Forgione, M. and Piga, D.},
year = {2020},
doi = {},
url = {https://arxiv.org/pdf/1911.13034.pdf}
}
DownloadForgione, M., Piga, D., Bemporad, A. (2020). Efficient Calibration of Embedded MPC. In Proceedings of the 21st IFAC World Congress (IFAC 20).
Efficient Calibration of Embedded MPC
Authors: Forgione, M. and Piga, D. and Bemporad, A.
Year: 2020
Abstract: Model Predictive Control (MPC) is a powerful and flexible design tool of high-performance controllers for physical systems in the presence of input and output constraints. A challenge for the practitioner applying MPC is the need of tuning a large number of parameters such as prediction and control horizons, weight matrices of the MPC cost function, and observer gains, according to different trade-offs. The MPC design task is even more involved when the control law has to be deployed to an embedded hardware unit endowed with limited computational resources. In this case, real-time system requirements limit the complexity of the applicable MPC configuration, engendering additional design tradeoffs and requiring to tune further parameters, such as the sampling time and the tolerances used in the on-line numerical solver. To take into account closed-loop performance and real-time requirements, in this paper we tackle the em-bedded MPC design problem using a global, data-driven, optimization approach We showcase the potential of this approach by tuning an MPC controller on two hardware platforms characterized by largely different computational capabilities
Published in Proceedings of the 21st IFAC World Congress (IFAC 20).
Efficient Calibration of Embedded MPC
@INPROCEEDINGS{forgione2020a,
title = {Efficient {C}alibration of {E}mbedded {MPC}},
booktitle = {Proceedings of the 21st {IFAC} World Congress ({IFAC} 20)},
author = {Forgione, M. and Piga, D. and Bemporad, A.},
year = {2020},
doi = {},
url = {https://arxiv.org/pdf/1911.13021.pdf}
}
Downloadvan der Gaag, L.C., Bolt, J.H. (2020). Poset representations for sets of elementary triplets. In Proceedings of the 10th International Conference on Probabilistic Graphical Models (PGM 2020), JMLR.org.
Poset representations for sets of elementary triplets
Authors: van der Gaag, L.C. and Bolt, J.H.
Year: 2020
Abstract: Semi-graphoid independence relations, composed of independence triplets, are typically exponentially large in the number of variables involved. For compact representation of such a relation, just a subset of its triplets, called a basis, are listed explicitly, while its other triplets remain implicit through a set of derivation rules. Two types of basis were defined for this purpose, which are the dominant-triplet basis and the elementary-triplet basis, of which the latter is commonly assumed to be significantly larger in size in general. In this paper we introduce the elementary po-triplet as a compact representation of multiple elementary triplets, by using separating posets. By exploiting this new representation, the size of an elementary-triplet basis can be reduced considerably. For computing the elementary closure of a starting set of po-triplets, we present an elegant algorithm that operates on the least and largest elements of the separating posets involved.
Published in Proceedings of the 10th International Conference on Probabilistic Graphical Models (PGM 2020), JMLR.org.
Poset representations for sets of elementary triplets
@INPROCEEDINGS{Linda2020c,
title = {Poset representations for sets of elementary triplets},
publisher = {JMLR.org},
booktitle = {Proceedings of the 10th International Conference on Probabilistic Graphical Models ({PGM} 2020)},
author = {van der Gaag, L.C. and Bolt, J.H.},
year = {2020},
doi = {},
url = {https://pgm2020.cs.aau.dk/index.php/accepted-papers/}
}
Downloadvan der Gaag, L.C., Renooij, S., Facchini, A. (2020). Building causal interaction models by recursive unfolding. In Proceedings of the 10th International Conference on Probabilistic Graphical Models (PGM 2020), JMLR.org.
Building causal interaction models by recursive unfolding
Authors: van der Gaag, L.C. and Renooij, S. and Facchini, A.
Year: 2020
Abstract: Causal interaction models, such as the well-known noisy-or and leaky noisy-or models, have become quite popular as a means to parameterize conditional probability tables for Bayesian networks. In this paper we focus on the engineering of subnetworks to represent such models and present a novel technique called recursive unfolding for this purpose. This technique allows inserting, removing and merging cause variables in an interaction model at will, without affecting the underlying represented information. We detail the technique, with the recursion invariants involved, and illustrate its practical use for Bayesian-network engineering by means of a small example.
Accepted in Proceedings of the 10th International Conference on Probabilistic Graphical Models (PGM 2020), JMLR.org.
Building causal interaction models by recursive unfolding
@INPROCEEDINGS{Linda2020a,
title = {Building causal interaction models by recursive unfolding},
publisher = {JMLR.org},
booktitle = {Proceedings of the 10th International Conference on Probabilistic Graphical Models ({PGM} 2020)},
author = {van der Gaag, L.C. and Renooij, S. and Facchini, A.},
year = {2020},
doi = {},
url = {https://pgm2020.cs.aau.dk/index.php/accepted-papers/}
}
DownloadGeh, R., Mauá, D.D., Antonucci, A. (2020). Learning probabilistic sentential decision diagrams by sampling. In Proceedings of the Eight Symposium on Knowledge Discovery, Mining and Learning (KMILE 2020), SBC, Porto Alegre, RS, Brasil, pp. 129–136.
Learning probabilistic sentential decision diagrams by sampling
Authors: Geh, R. and Mauá, D.D. and Antonucci, A.
Year: 2020
Abstract: Probabilistic circuits are deep probabilistic models with neural-network-like semantics capable of accurately and efficiently answering probabilistic queries without sacrificing expressiveness. Probabilistic Sentential Decision Diagrams (PSDDs) are a subclass of probabilistic circuits able to embed logical constraints to the circuit’s structure. In doing so, they obtain extra expressiveness with empirical optimal performance. Despite achieving competitive performance compared to other state-of-the-art competitors, there have been very few attempts at learning PSDDs from a combination of both data and knowledge in the form of logical formulae. Our work investigates sampling random PSDDs consistent with domain knowledge and evaluating against state-of-the-art probabilistic models. We propose a method of sampling that retains important structural constraints on the circuit’s graph that guarantee query tractability. Finally, we show that these samples are able to achieve competitive performance even on larger domains.
Published in Proceedings of the Eight Symposium on Knowledge Discovery, Mining and Learning (KMILE 2020), SBC, Porto Alegre, RS, Brasil, pp. 129–136.
Learning probabilistic sentential decision diagrams by sampling
@INPROCEEDINGS{antonucci2020c,
title = {Learning probabilistic sentential decision diagrams by sampling},
publisher = {SBC},
address = {Porto Alegre, RS, Brasil},
booktitle = {Proceedings of the Eight Symposium on Knowledge Discovery, Mining and Learning ({KMILE} 2020)},
author = {Geh, R. and Mau\'a, D.D. and Antonucci, A.},
pages = {129--136},
year = {2020},
doi = {10.5753/kdmile.2020.11968},
url = {}
}
DownloadHuber, D., Cabañas, R., Antonucci, A., Zaffalon, M. (2020). CREMA: a Java library for credal network inference. In Jaeger, M., Nielsen, T.D. (Eds), Proceedings of the 10th International Conference on Probabilistic Graphical Models (PGM 2020), Proceedings of Machine Learning Research, PMLR, Aalborg, Denmark.
CREMA: a Java library for credal network inference
Authors: Huber, D. and Cabañas, R. and Antonucci, A. and Zaffalon, M.
Year: 2020
Abstract: We present CREMA (Credal Models Algorithms), a Java library for inference in credal networks. These models are analogous to Bayesian networks, but their local parameters are only constrained to vary in, so-called credal, sets. Inference in credal networks is intended as the computation of the bounds of a query with respect to those local variations. For credal networks the task is harder than in Bayesian networks, being NPPP-hard in general models. Yet, scalable approximate algorithms have been shown to provide good accuracies on large or dense models, while exact techniques can be designed to process small or sparse models. CREMA embeds these algorithms and also offers an API to build and query credal networks together with a specification format. This makes CREMA, whose features are discussed and described by a simple example, the most advanced tool for credal network modelling and inference developed so far.
Published in Jaeger, M., Nielsen, T.D. (Eds), Proceedings of the 10th International Conference on Probabilistic Graphical Models (PGM 2020), Proceedings of Machine Learning Research, PMLR, Aalborg, Denmark.
CREMA: a Java library for credal network inference
@INPROCEEDINGS{huber2020a,
title = {{CREMA}: a {J}ava library for credal network inference},
editor = {Jaeger, M. and Nielsen, T.D.},
publisher = {PMLR},
address = {Aalborg, Denmark},
series = {Proceedings of Machine Learning Research},
booktitle = {Proceedings of the 10th International Conference on Probabilistic Graphical Models ({PGM} 2020)},
author = {Huber, D. and Caba\~nas, R. and Antonucci, A. and Zaffalon, M.},
year = {2020},
doi = {},
url = {https://pgm2020.cs.aau.dk}
}
DownloadKanjirangat, V., Mellace, S., Antonucci, A. (2020). Temporal embeddings and transformer models for narrative text understanding. In Third International Workshop on Narrative Extraction from Texts (Text2Story 20), 42nd European Conference on Information Retrieval (ECIR 20), ceur.
Temporal embeddings and transformer models for narrative text understanding
Authors: Kanjirangat, V. and Mellace, S. and Antonucci, A.
Year: 2020
Abstract: We present two deep learning approaches to narrative text understanding for character relationship modelling. The temporal evolution of these relations is described by dynamic word embeddings, that are designed to learn semantic changes over time. An empirical analysis of the corresponding character trajectories shows that such approaches are effective in depicting dynamic evolution. A supervised learning approach based on the state-of-the-art transformer model BERT is used instead to detect static relations between characters. The empirical validation shows that such events (e.g., two characters belonging to the same family) might be spotted with good accuracy, even when using automatically annotated data. This provides a deeper understanding of narrative plots based on the identification of key facts. Standard clustering techniques are finally used for character de-aliasing, a necessary pre-processing step for both approaches. Overall, deep learning models appear to be suitable for narrative text understanding, while also providing a challenging and unexploited benchmark for general natural language understanding.
Published in Third International Workshop on Narrative Extraction from Texts (Text2Story 20), 42nd European Conference on Information Retrieval (ECIR 20), ceur.
Temporal embeddings and transformer models for narrative text understanding
@INPROCEEDINGS{vani2020b,
title = {Temporal embeddings and transformer models for narrative text understanding},
publisher = {ceur},
booktitle = {Third International Workshop on Narrative Extraction {f}rom Texts ({Text2Story} 20), 42nd European Conference on Information Retrieval ({ECIR} 20)},
author = {Kanjirangat, V. and Mellace, S. and Antonucci, A.},
year = {2020},
doi = {},
url = {http://ceur-ws.org/Vol-2593/paper9.pdf}
}
DownloadKanjirangat, V., Mitrovic, S., Antonucci, A., Rinaldi, F. (2020). SST-BERT at SemEval-2020 task 1: semantic shift tracing by clustering in BERT-based embedding spaces. In SemEval-2020 Task 1: Unsupervised Lexical Semantic Change Detection.To appear In Proceedings of the 14th International Workshop on Semantic Evaluation, Barcelona, Spain.
SST-BERT at SemEval-2020 task 1: semantic shift tracing by clustering in BERT-based embedding spaces
Authors: Kanjirangat, V. and Mitrovic, S. and Antonucci, A. and Rinaldi, F.
Year: 2020
Abstract: Lexical semantic change detection (also known as semantic shift tracing) is a task of identifying words that have changed their meaning over time. Unsupervised semantic shift tracing, focal point of SemEval2020, is particularly challenging. Given the unsupervised setup, in this work, we propose to identify clusters among different occurrences of each target word, considering these as representatives of different word meanings. As such, disagreements in obtained clusters naturally allow to quantify the level of semantic shift per each target word in four target languages. To leverage this idea, clustering is performed on contextualized (BERT-based) embeddings of word occurrences. The obtained results show that our approach performs well both measured separately (per language) and overall, where we surpass all provided SemEval baselines.
Published in SemEval-2020 Task 1: Unsupervised Lexical Semantic Change Detection.To appear In Proceedings of the 14th International Workshop on Semantic Evaluation, Barcelona, Spain.
SST-BERT at SemEval-2020 task 1: semantic shift tracing by clustering in BERT-based embedding spaces
@INPROCEEDINGS{vani2020semeval,
title = {{SST}-{BERT} at {SemEval}-2020 task 1: semantic shift tracing by clustering in {BERT}-based embedding spaces},
booktitle = {{SemEval}-2020 Task 1: Unsupervised Lexical Semantic Change Detection.To {a}ppear In Proceedings of the 14th International Workshop on Semantic Evaluation, Barcelona, Spain},
author = {Kanjirangat, V. and Mitrovic, S. and Antonucci, A. and Rinaldi, F.},
year = {2020},
doi = {},
url = {http://alt.qcri.org/semeval2020/}
}
DownloadKern, H., Corani, G., Huber, D., Vermes, N., Zaffalon, M., Varini, M., Wenzel, C., Fringer, A. (2020). Impact on place of death in cancer patients: a causal exploration in southern switzerland. BMC Palliative Care 19, 160.
Impact on place of death in cancer patients: a causal exploration in southern switzerland
Authors: Kern, H. and Corani, G. and Huber, D. and Vermes, N. and Zaffalon, M. and Varini, M. and Wenzel, C. and Fringer, A.
Year: 2020
Abstract:
Background
Most terminally ill cancer patients prefer to die at home, but a majority die in institutional settings. Research questions about this discrepancy have not been fully answered. This study applies artificial intelligence and machine learning techniques to explore the complex network of factors and the cause-effect relationships affecting the place of death, with the ultimate aim of developing policies favouring home-based end-of-life care.
Methods
A data mining algorithm and a causal probabilistic model for data analysis were developed with information derived from expert knowledge that was merged with data from 116 deceased cancer patients in southern Switzerland. This data set was obtained via a retrospective clinical chart review.
Results
Dependencies of disease and treatment-related decisions demonstrate an influence on the place of death of 13%. Anticancer treatment in advanced disease prevents or delays communication about the end of life between oncologists, patients and families. Unknown preferences for the place of death represent a great barrier to a home death. A further barrier is the limited availability of family caregivers for terminal home care. The family’s preference for the last place of care has a high impact on the place of death of 51%, while the influence of the patient’s preference is low, at 14%. Approximately one-third of family systems can be empowered by health care professionals to provide home care through open end-of-life communication and good symptom management. Such intervention has an influence on the place of death of 17%. If families express a convincing preference for home care, the involvement of a specialist palliative home care service can increase the probability of home deaths by 24%.
Conclusion
Concerning death at home, open communication about death and dying is essential. Furthermore, for the patient preference for home care to be respected, the family’s decision for the last place of care seems to be key. The early initiation of family-centred palliative care and the provision of specialist palliative home care for patients who wish to die at home are suggested.
Published in BMC Palliative Care 19, Research Square, 160.
Impact on place of death in cancer patients: a causal exploration in southern switzerland
@ARTICLE{Kern2020,
title = {Impact on place of death in cancer patients: a causal exploration in southern switzerland},
journal = {{BMC} Palliative Care},
publisher = {Research Square},
volume = {19},
author = {Kern, H. and Corani, G. and Huber, D. and Vermes, N. and Zaffalon, M. and Varini, M. and Wenzel, C. and Fringer, A.},
pages = {160},
year = {2020},
doi = {10.21203/rs.3.rs-29758/v3},
url = {}
}
DownloadLaurain, V., Tóth, R., Piga, D., Darwish, M.A.H. (2020). Sparse RKHS estimation via globally convex optimization and its application in LPV-IO identification. Automatica 115.
Sparse RKHS estimation via globally convex optimization and its application in LPV-IO identification
Authors: Laurain, V. and Tóth, R. and Piga, D. and Darwish, M.A.H.
Year: 2020
Abstract: Reproducing kernel Hilbert spaces, Elastic net, Support vector machines, Gaussian processes, Non-parametric estimation, Linear parameter-varying systems, Model order selection",
abstract = "Function estimation using the Reproducing Kernel Hilbert Space (RKHS) framework is a powerful tool for identification of a general class of nonlinear dynamical systems without requiring much a priori information on model orders and nonlinearities involved. However, the high degrees-of-freedom (DOFs) of RKHS estimators has its price, as in case of large scale function estimation problems, they often require a serious amount of data samples to explore the search space adequately for providing high-performance model estimates. In cases where nonlinear dynamic relations can be expressed as a sum of functions, the literature proposes solutions to this issue by enforcing sparsity for adequate restriction of the DOFs of the estimator, resulting in parsimonious model estimates. Unfortunately, all existing solutions are based on greedy approaches, leading to optimization schemes which cannot guarantee convergence to the global optimum. In this paper, we propose an L1-regularized non-parametric RKHS estimator which is the solution of a quadratic optimization problem. Effectiveness of the scheme is demonstrated on the non-parametric identification problem of LPV-IO models where the method solves simultaneously (i) the model order selection problem (in terms of number of input–output lags and input delay in the model structure) and (ii) determining the unknown functional dependency of the model coefficients on the scheduling variable directly from data. The paper also provides an extensive simulation study to illustrate effectiveness of the proposed scheme.
Published in Automatica 115.
Sparse RKHS estimation via globally convex optimization and its application in LPV-IO identification
@ARTICLE{piga2020b,
title = {Sparse {RKHS} estimation via globally convex optimization and its application in {LPV}-{IO} identification},
journal = {Automatica},
volume = {115},
author = {Laurain, V. and T\'oth, R. and Piga, D. and Darwish, M.A.H.},
year = {2020},
doi = {10.1016/j.automatica.2020.108914},
url = {http://www.sciencedirect.com/science/article/pii/S0005109820301126}
}
DownloadLiew, B.X.W., Peolsson, A., Scutari, M., Löfgren, H., Wibault, J., r A Dedering, , Öberg, B., Zsigmond, P., Falla, D. (2020). Probing the mechanisms underpinning recovery in post-surgical patients with cervical radiculopathy using bayesian networks. European Journal of Pain 24(5), pp. 909–920.
Probing the mechanisms underpinning recovery in post-surgical patients with cervical radiculopathy using bayesian networks
Authors: Liew, B.X.W. and Peolsson, A. and Scutari, M. and Löfgren, H. and Wibault, J. and r A Dedering, and Öberg, B. and Zsigmond, P. and Falla, D.
Year: 2020
Published in European Journal of Pain 24(5), pp. 909–920.
Probing the mechanisms underpinning recovery in post-surgical patients with cervical radiculopathy using bayesian networks
@ARTICLE{scutari20a,
title = {Probing the mechanisms underpinning recovery in post-surgical patients with cervical radiculopathy using bayesian networks},
journal = {European Journal of Pain},
volume = {24},
author = {Liew, B.X.W. and Peolsson, A. and Scutari, M. and L\"ofgren, H. and Wibault, J. and r A Dedering, and \"Oberg, B. and Zsigmond, P. and Falla, D.},
number = {5},
pages = {909--920},
year = {2020},
doi = {},
url = {}
}
DownloadMangili, F., Broggini, D., Antonucci, A. (2020). Conversational recommender system by Bayesian methods. In Davis, Jesse, Tabia, Karim (Eds), Proceedings of the Fourteenth International Conference on Scalable Uncertainty Management (SUM 2020), Lecture Notes in Artificial Intelligence 12322, Springer, Cham, pp. 200–213.
Conversational recommender system by Bayesian methods
Authors: Mangili, F. and Broggini, D. and Antonucci, A.
Year: 2020
Abstract: We present a Bayesian approach to conversational recommender systems. After any interaction with the user, a probability mass function over the items is updated by the system. The conversational feature corresponds to a sequential discovery of the user preferences based on questions. Information-theoretic criteria are used to optimally shape the interactions and decide when the conversation ends. Most probable items are consequently recommended. Dedicated elicitation techniques for the prior probabilities of the parameters modelling the interactions are derived from basic structural judgements based on logical compatibility and symmetry assumptions. Such prior knowledge is combined with data for better item discrimination. Our Bayesian approach is validated against matrix factorization techniques for cold-start recommendations based on metadata using the popular benchmark data set MovieLens. Results show that the proposed approach allows to considerably reduce the number of interactions while maintaining good ranking performance.
Published in Davis, Jesse, Tabia, Karim (Eds), Proceedings of the Fourteenth International Conference on Scalable Uncertainty Management (SUM 2020), Lecture Notes in Artificial Intelligence 12322, Springer, Cham, pp. 200–213.
Conversational recommender system by Bayesian methods
@INPROCEEDINGS{mangili2020b,
title = {Conversational recommender system by {B}ayesian methods},
editor = {Davis, Jesse and Tabia, Karim},
publisher = {Springer, Cham},
series = {Lecture Notes in Artificial Intelligence},
volume = {12322},
booktitle = {Proceedings of the Fourteenth International Conference on Scalable Uncertainty Management ({SUM} 2020)},
author = {Mangili, F. and Broggini, D. and Antonucci, A.},
pages = {200--213},
year = {2020},
doi = {10.1007/978-3-030-58449-8_14},
url = {}
}
DownloadMangili, F., Broggini, D., Antonucci, A., Alberti, M., Cimasoni, L. (2020). A Bayesian approach to conversational recommendation systems. AAAI 2020 Workshop on Interactive and Conversational Recommendation Systems (WICRS-20).
A Bayesian approach to conversational recommendation systems
Authors: Mangili, F. and Broggini, D. and Antonucci, A. and Alberti, M. and Cimasoni, L.
Year: 2020
Abstract: We present a conversational recommendation system based on a Bayesian approach. A probability mass function over the items is updated after any interaction with the user, with information-theoretic criteria optimally shaping the interaction and deciding when the conversation should be terminated and the most probable item consequently recommended. Dedicated elicitation techniques for the prior probabilities of the parameters modeling the interactions are derived from basic structural judgements. Such prior information can be combined with historical data to discriminate items with different recommendation histories. A case study based on the application of this approach to stagend.com, an online platform for booking entertainers, is finally discussed together with an empirical analysis showing the advantages in terms of recommendation quality and efficiency.
Accepted in AAAI 2020 Workshop on Interactive and Conversational Recommendation Systems (WICRS-20).
Note: Accepted for oral presentation
A Bayesian approach to conversational recommendation systems
@ARTICLE{mangili2020a,
title = {A {B}ayesian approach to conversational recommendation systems},
journal = {{AAAI} 2020 Workshop on Interactive and Conversational Recommendation Systems ({WICRS}-20)},
author = {Mangili, F. and Broggini, D. and Antonucci, A. and Alberti, M. and Cimasoni, L.},
year = {2020},
doi = {},
url = {https://sites.google.com/view/wicrs2020}
}
DownloadMattei, L., Antonucci, A., Mauá, D.D., Facchini, A., Llerena, J.V. (2020). Tractable inference in credal sentential decision diagrams. International Journal of Approximate Reasoning 125, pp. 26–48.
Tractable inference in credal sentential decision diagrams
Authors: Mattei, L. and Antonucci, A. and Mauá, D.D. and Facchini, A. and Llerena, J.V.
Year: 2020
Abstract: Probabilistic sentential decision diagrams are logic circuits where the inputs of disjunctive gates are annotated by probability values. They allow for a compact representation of joint probability mass functions defined over sets of Boolean variables, that are also consistent with the logical constraints defined by the circuit. The probabilities in such a model are usually “learned” from a set of observations. This leads to overconfident and prior-dependent inferences when data are scarce, unreliable or conflicting. In this work, we develop the credal sentential decision diagrams, a generalisation of their probabilistic counterpart that allows for replacing the local probabilities with (so-called credal) sets of mass functions. These models induce a joint credal set over the set of Boolean variables, that sharply assigns probability zero to states inconsistent with the logical constraints. Three inference algorithms are derived for these models. These allow to compute: (i) the lower and upper probabilities of an observation for an arbitrary number of variables; (ii) the lower and upper conditional probabilities for the state of a single variable given an observation; (iii) whether or not all the probabilistic sentential decision diagrams compatible with the credal specification have the same most probable explanation of a given set of variables given an observation of the other variables. These inferences are tractable, as all the three algorithms, based on bottom-up traversal with local linear programming tasks on the disjunctive gates, can be solved in polynomial time with respect to the circuit size. The first algorithm is always exact, while the remaining two might induce a conservative (outer) approximation in the case of multiply connected circuits. A semantics for this approximation together with an auxiliary algorithm able to decide whether or not the result is exact is also provided together with a brute-force characterization of the exact inference in these cases. For a first empirical validation, we consider a simple application based on noisy seven-segment display images. The credal models are observed to properly distinguish between easy and hard-to-detect instances and outperform other generative models not able to cope with logical constraints.
Published in International Journal of Approximate Reasoning 125, pp. 26–48.
Tractable inference in credal sentential decision diagrams
@ARTICLE{antonucci2020d,
title = {Tractable inference in credal sentential decision diagrams},
journal = {International Journal of Approximate Reasoning},
volume = {125},
author = {Mattei, L. and Antonucci, A. and Mau\'a, D.D. and Facchini, A. and Llerena, J.V.},
pages = {26--48},
year = {2020},
doi = {10.1016/j.ijar.2020.06.005},
url = {}
}
DownloadMauà, D.D., Ribeiro, H., Katague, G., Antonucci, A. (2020). Two reformulation approaches to maximum-a-posteriori inference in sum-product networks. In Jaeger, M., Nielsen, T.D. (Eds), Proceedings of the 10th International Conference on Probabilistic Graphical Models (PGM 2020), Proceedings of Machine Learning Research, PMLR, Aalborg, Denmark.
Two reformulation approaches to maximum-a-posteriori inference in sum-product networks
Authors: Mauà, D.D. and Ribeiro, H. and Katague, G. and Antonucci, A.
Year: 2020
Abstract: There exists a dichotomy between classical probabilistic graphical models, such as Bayesian net- works (BNs), and modern tractable models, such as sum-product networks (SPNs). The former generally have intractable inference, but provide a high level of interpretability, while the latter admit a wide range of tractable inference routines, but are typically harder to interpret. Due to this dichotomy, tools to convert between BNs and SPNs are desirable. While one direction – compiling BNs into SPNs – is well discussed in Darwiche’s seminal work on arithmetic circuit compilation, the converse direction – decompiling SPNs into BNs – has received surprisingly little attention. In this paper, we fill this gap by proposing SPN2BN, an algorithm that decompiles an SPN into a BN. SPN2BN has several salient features when compared to the only other two works decompiling SPNs. Most significantly, the BNs returned by SPN2BN are minimal independence-maps that are more parsimonious with respect to the introduction of latent variables. Secondly, the output BN produced by SPN2BN can be precisely characterized with respect to a compiled BN. More specifically, a certain set of directed edges will be added to the input BN, giving what we will call the moral-closure. Lastly, it is established that our compilation-decompilation process is idempotent. This has practical significance as it limits the size of the decompiled SPN.
Published in Jaeger, M., Nielsen, T.D. (Eds), Proceedings of the 10th International Conference on Probabilistic Graphical Models (PGM 2020), Proceedings of Machine Learning Research, PMLR, Aalborg, Denmark.
Two reformulation approaches to maximum-a-posteriori inference in sum-product networks
@INPROCEEDINGS{antonucci2020b,
title = {Two reformulation approaches to maximum-a-posteriori inference in sum-product networks},
editor = {Jaeger, M. and Nielsen, T.D.},
publisher = {PMLR},
address = {Aalborg, Denmark},
series = {Proceedings of Machine Learning Research},
booktitle = {Proceedings of the 10th International Conference on Probabilistic Graphical Models ({PGM} 2020)},
author = {Mau\`a, D.D. and Ribeiro, H. and Katague, G. and Antonucci, A.},
year = {2020},
doi = {},
url = {https://pgm2020.cs.aau.dk}
}
DownloadMejari, M., Breschi, V., Naik, V.V., Piga, D. (2020). A bias-correction approach for the identification of piecewise affine output-error models. In 21st IFAC World Congress (IFAC 2020), Berlin, Germany.
A bias-correction approach for the identification of piecewise affine output-error models
Authors: Mejari, M. and Breschi, V. and Naik, V.V. and Piga, D.
Year: 2020
Abstract: The paper presents an algorithm for the identification of PieceWise Affine Output-Error (PWA-OE) models, which involves the estimation of the parameters defining affine submodels as well as a partition of the regressor space. For the estimation of affine submodel parameters, a bias-correction scheme is presented to correct the bias in the least-squares estimates which is caused by the output-error noise structure. The obtained bias-corrected estimates are proven to be consistent under suitable assumptions. The bias-correction method is then combined with a recursive estimation algorithm for clustering the regressors.
These clusters are used to compute a partition of the regressor space by employing linear multi-category discrimination. The effectiveness of the proposed methodology is demonstrated via a simulation case study.
Accepted in 21st IFAC World Congress (IFAC 2020), Berlin, Germany.
A bias-correction approach for the identification of piecewise affine output-error models
@INPROCEEDINGS{mejari2020a,
title = {A bias-correction approach for the identification of piecewise affine output-error models},
address = {Berlin, Germany},
booktitle = {21st {IFAC} World Congress ({IFAC} 2020)},
author = {Mejari, M. and Breschi, V. and Naik, V.V. and Piga, D.},
year = {2020},
doi = {},
url = {https://www.ifac2020.org}
}
DownloadMejari, M., Breschi, V., Piga, D. (2020). Recursive bias-correction method for identification of piecewise affine output-error models. IEEE Control Systems Letters 4, pp. 970–975.
Recursive bias-correction method for identification of piecewise affine output-error models
Authors: Mejari, M. and Breschi, V. and Piga, D.
Year: 2020
Abstract: Learning PieceWise Affine Output-Error (PWA-OE) models from data requires to estimate a finite set of affine output-error sub-models as well as a partition of the regressors space over which the sub-models are defined. For an output-error type noise structure, the algorithms based on ordinary least squares (LS) fail to compute a consistent estimate of the sub-model parameters. On the other hand, the prediction error methods (PEMs) provide a consistent parameter estimate, however, they require to solve a non-convex optimization problem for which the numerical algorithms may get trapped in a local minimum, leading to inaccurate estimates. In this paper, we propose a recursive bias-correction scheme for identifying PWA-OE models, retaining the computational efficiency of the standard LS algorithms while providing a consistent estimate of the sub-model parameters, under suitable assumptions. The proposed approach allows one to recursively update the estimates of the sub-models parameters and to cluster the regressors. Linear multi-category techniques are then employed to estimate a partition of the regressor space based on the estimated clusters. The performance of the proposed algorithm is demonstrated via an academic example.
Published in IEEE Control Systems Letters 4, pp. 970–975.
Recursive bias-correction method for identification of piecewise affine output-error models
@ARTICLE{mejari2020b,
title = {Recursive bias-correction method for identification of piecewise affine output-error models},
journal = {{IEEE} Control Systems Letters},
volume = {4},
author = {Mejari, M. and Breschi, V. and Piga, D.},
pages = {970--975},
year = {2020},
doi = {10.1109/LCSYS.2020.2998282},
url = {}
}
DownloadMiranda, E., Zaffalon, M. (2020). Compatibility, desirability, and the running intersection property. Artificial Intelligence 283, 103724.
Compatibility, desirability, and the running intersection property
Authors: Miranda, E., Zaffalon, M.
Year: 2020
Abstract: Compatibility is the problem of checking whether some given probabilistic assessments have a common joint probabilistic model. When the assessments are unconditional, the problem is well established in the literature and finds a solution through the running intersection property (RIP). This is not the case of conditional assessments. In this paper, we study the compatibility problem in a very general setting: any possibility space, unrestricted domains, imprecise (and possibly degenerate) probabilities. We extend the unconditional case to our setting, thus generalising most of previous results in the literature. The conditional case turns out to be fundamentally different from the unconditional one. For such a case, we prove that the problem can still be solved in general by RIP but in a more involved way: by constructing a junction tree and propagating information over it. Still, RIP does not allow us to optimally take advantage of sparsity: in fact, conditional compatibility can be simplified further by joining junction trees with coherence graphs.
Published in Artificial Intelligence 283, 103724.
Compatibility, desirability, and the running intersection property
@ARTICLE{zaffalon2020a,
title = {Compatibility, desirability, and the running intersection property},
journal = {Artificial Intelligence},
volume = {283},
author = {Miranda, E., Zaffalon, M.},
pages = {103724},
year = {2020},
doi = {10.1016/j.artint.2020.103274},
url = {}
}
DownloadPiga, D., Bemporad, A., Benavoli, A. (2020). Rao-Blackwellized sampling for batch and recursive Bayesian inference of piecewise affine models. Automatica 117, 109002.
Rao-Blackwellized sampling for batch and recursive Bayesian inference of piecewise affine models
Authors: Piga, D. and Bemporad, A. and Benavoli, A.
Year: 2020
Abstract: This paper addresses batch (offline) and recursive (online) Bayesian inference of Piecewise Affine (PWA) regression models. By exploiting the particular structure of PWA models, efficient Rao-Blackwellized Monte Carlo sampling algorithms are developed to approximate the joint posterior distribution of the model parameters. Only the marginal posterior of the parameters used to describe the regressor-space partition is approximated, either in a batch mode using a Metropolis–Hastings Markov-Chain Monte Carlo (MCMC) sampler, or sequentially using particle filters, while the conditional distribution of the other model parameters is computed analytically. Probability distributions for the predicted outputs given new test inputs are derived and modifications of the proposed approaches to address maximum-a-posteriori estimate are discussed. The performance of the proposed algorithms is shown via a numerical example and through a benchmark case study on data-driven modelling of the electronic component placement process in a pick-and-place machine.
Published in Automatica 117, 109002.
Rao-Blackwellized sampling for batch and recursive Bayesian inference of piecewise affine models
@ARTICLE{piga2020a,
title = {Rao-{B}lackwellized sampling for batch and recursive {B}ayesian inference of piecewise affine models},
journal = {Automatica},
volume = {117},
author = {Piga, D. and Bemporad, A. and Benavoli, A.},
pages = {109002},
year = {2020},
doi = {10.1016/j.automatica.2020.109002},
url = {}
}
DownloadPiga, D., Breschi, V., Bemporad, A. (2020). Estimation of jump box–jenkins models. Automatica 120, 109126.
Estimation of jump box–jenkins models
Authors: Piga, D. and Breschi, V. and Bemporad, A.
Year: 2020
Abstract: Jump Box–Jenkins (BJ) models are a collection of a finite set of linear dynamical submodels in BJ form that switch over time, according to a Markov chain. This paper addresses the problem of maximum-a-posteriori estimation of jump BJ models from a given training input/output dataset. The proposed solution method estimates the coefficients of the BJ submodels, the state transition probabilities of the Markov chain regulating the switching of operating modes, and the corresponding mode sequence hidden in the dataset. In particular, the posterior distribution of all the unknown variables characterizing the jump BJ model is derived and then maximized using a coordinate ascent algorithm. The resulting estimation algorithm alternates between Gauss–Newton optimization of the coefficients of the BJ submodels, a method derived based on an instance of prediction error methods tailored to BJ models with switching coefficients, and approximated dynamic programming for optimization of the sequence of active modes. The quality of the proposed estimation approach is evaluated on a numerical example based on synthetic data and in a case study related to segmentation of honeybee dances.
Published in Automatica 120, 109126.
Estimation of jump box–jenkins models
@ARTICLE{piga2020c,
title = {Estimation of jump box--jenkins models},
journal = {Automatica},
volume = {120},
author = {Piga, D. and Breschi, V. and Bemporad, A.},
pages = {109126},
year = {2020},
doi = {10.1016/j.automatica.2020.109126},
url = {http://www.sciencedirect.com/science/article/pii/S0005109820303241}
}
DownloadRoveda, L., Bussolan, A., Braghin, F., Piga, D. (2020). 6D virtual sensor for wrench estimation in robotized interaction tasks exploiting extended Kalman filter. MDPI Machines.
6D virtual sensor for wrench estimation in robotized interaction tasks exploiting extended Kalman filter
Authors: Roveda, L. and Bussolan, A. and Braghin, F. and Piga, D.
Year: 2020
Abstract: Industrial robots are commonly used to perform interaction tasks (such as assemblies or
polishing), requiring the robot to be in contact with the surrounding environment. Such environments
are (partially) unknown to the robot controller. Therefore, there is the need to implement interaction
controllers capable of suitably reacting to the established contacts. Although standard force controllers
require force/torque measurements to close the loop, most of the industrial manipulators do not have
installed force/torque sensor(s). In addition, the integration of external sensors results in additional
costs and implementation effort, not affordable in many contexts/applications. To extend the use
of compliant controllers to sensorless interaction control, a model-based methodology is presented
in this paper for the online estimation of the interaction wrench, implementing a 6D virtual sensor.
Relying on sensorless Cartesian impedance control, an Extended Kalman Filter (EKF) is proposed
for the interaction wrench estimation. The described approach has been validated in simulations,
taking into account four different scenarios. In addition, experimental validation has been performed
employing a Franka EMIKA panda robot. A human–robot interaction scenario and an assembly task
have been considered to show the capabilities of the developed EKF, which is able to perform the
estimation with high bandwidth, achieving convergence with limited errors.
Published in MDPI Machines.
6D virtual sensor for wrench estimation in robotized interaction tasks exploiting extended Kalman filter
@ARTICLE{Roveda2020b,
title = {{6D} virtual sensor for wrench estimation in robotized interaction tasks exploiting extended {K}alman filter},
journal = {{MDPI} Machines},
author = {Roveda, L. and Bussolan, A. and Braghin, F. and Piga, D.},
year = {2020},
doi = {10.3390/machines8040067},
url = {}
}
DownloadRoveda, L., Castaman, N., Franceschi, P., Ghidoni, S., Pedrocchi, N. (2020). A control framework definition to overcome position/interaction dynamics uncertainties in force-controlled tasks. In IEEE International Conference on Robotics and Automation (ICRA) 2020.
A control framework definition to overcome position/interaction dynamics uncertainties in force-controlled tasks
Authors: Roveda, L. and Castaman, N. and Franceschi, P. and Ghidoni, S. and Pedrocchi, N.
Year: 2020
Abstract: Within the Industry 4.0 context, industrial robots need to show increasing autonomy. The manipulator has to be able to react to uncertainties/changes in the working environment, displaying a robust behavior. In this paper, a control framework is proposed to perform industrial interaction tasks in uncertain working scenes. The proposed methodology relies
on two components: i) a 6D pose estimation algorithm aiming
to recognize large and featureless parts; ii) a variable damping impedance controller (inner loop) enhanced by an adaptive saturation PI (outer loop) for high accuracy force control (i.e., zero steady-state force error and force overshoots avoidance). The proposed methodology allows to be robust w.r.t. task uncertainties (i.e., positioning errors and interaction dynamics). The proposed approach has been evaluated in an assembly task of a side-wall panel to be installed inside the aircraft cabin. As a test platform, the KUKA iiwa 14 R820 has been used together with the Microsoft Kinect 2.0 as RGB-D sensor. Experiments show the reliability in the 6D pose estimation and the high-performance in the force-tracking task, avoiding force overshoots while achieving the tracking of the reference force.
Published in IEEE International Conference on Robotics and Automation (ICRA) 2020.
A control framework definition to overcome position/interaction dynamics uncertainties in force-controlled tasks
@INPROCEEDINGS{Roveda2020f,
title = {A control framework definition to overcome position/interaction dynamics uncertainties in force-controlled tasks},
booktitle = {{IEEE} International Conference on Robotics and Automation ({ICRA}) 2020},
author = {Roveda, L. and Castaman, N. and Franceschi, P. and Ghidoni, S. and Pedrocchi, N.},
year = {2020},
doi = {10.1109/ICRA40945.2020.9197141},
url = {}
}
DownloadRoveda, L., Forgione, M., Piga, D. (2020). Robot control parameters auto-tuning in trajectory tracking applications. Control Engineering Practice 101, 104488.
Robot control parameters auto-tuning in trajectory tracking applications
Authors: Roveda, L. and Forgione, M. and Piga, D.
Year: 2020
Abstract: Autonomy is increasingly demanded to industrial manipulators. Robots have to be capable to regulate their behavior to different operational conditions, adapting to the specific task to be executed without requiring high time/resource-consuming human intervention. Achieving an automated tuning of the control parameters of a manipulator is still a challenging task, which involves modeling/identification of the robot dynamics. This usually results in an onerous procedure, both in terms of experimental and data-processing time. This paper addresses the problem of automated tuning of the manipulator controller for trajectory tracking, optimizing control parameters based on the specific trajectory to be executed. A Bayesian optimization algorithm is proposed to tune both the low-level controller parameters (i.e., the equivalent link-masses of the feedback linearizator and the feedforward controller) and the high-level controller parameters (i.e., the joint PID gains). The algorithm adapts the control parameters through a data-driven procedure, optimizing a user-defined trajectory-tracking cost. Safety constraints ensuring, e.g., closed-loop stability and bounds on the maximum joint position error are also included. The performance of proposed approach is demonstrated on a torque-controlled 7-degree-of-freedom FRANKA Emika robot manipulator. The 25 robot control parameters (i.e., 4 link-mass parameters and 21 PID gains) are tuned in less than 130 iterations, and comparable results with respect to the FRANKA Emika embedded position controller are achieved. In addition, the generalization capabilities of the proposed approach are shown exploiting the proper reference trajectory for the tuning of the control parameters.
Published in Control Engineering Practice 101, 104488.
Robot control parameters auto-tuning in trajectory tracking applications
@ARTICLE{roveda2020a,
title = {Robot control parameters auto-tuning in trajectory tracking applications},
journal = {Control Engineering Practice},
volume = {101},
author = {Roveda, L. and Forgione, M. and Piga, D.},
pages = {104488},
year = {2020},
doi = {10.1016/j.conengprac.2020.104488},
url = {}
}
DownloadRoveda, L., Forgione, M., Piga, D. (2020). One-stage auto-tuning procedure of robot dynamics and control parameters for trajectory tracking applications. In Ubiquitous Robots 2020.
One-stage auto-tuning procedure of robot dynamics and control parameters for trajectory tracking applications
Authors: Roveda, L. and Forgione, M. and Piga, D.
Year: 2020
Abstract: Autonomy is increasingly demanded by industrial
manipulators. Robots have to be capable to regulate their
behavior to different operational conditions, without requiring high time/resource-consuming human intervention. Achieving an automated tuning of the control parameters of a manipulator is still a challenging task. This paper addresses the problem of automated tuning of the manipulator controller for trajectory tracking. A Bayesian optimization algorithm is proposed to tune both the low-level controller parameters (i.e., robot dynamics compensation) and the high-level controller parameters (i.e., the joint PID gains). The algorithm adapts the control parameters through a data-driven procedure, optimizing a user-defined trajectory-tracking cost. Safety constraints ensuring, e.g., closed-loop stability and bounds on the maximum joint position errors, are also included. The performance of the proposed approach is demonstrated on a torque-controlled 7-degree-of-freedom FRANKA Emika robot manipulator. The 25 robot control parameters (i.e., 4 link-mass parameters and 21 PID gains) are tuned in 125 iterations, and comparable results with respect to the FRANKA Emika embedded position controller are achieved.
Published in Ubiquitous Robots 2020.
One-stage auto-tuning procedure of robot dynamics and control parameters for trajectory tracking applications
@INPROCEEDINGS{Roveda2020h,
title = {One-stage auto-tuning procedure of robot dynamics and control parameters for trajectory tracking applications},
booktitle = {Ubiquitous Robots 2020},
author = {Roveda, L. and Forgione, M. and Piga, D.},
year = {2020},
doi = {10.1109/UR49135.2020.9144761},
url = {}
}
DownloadRoveda, L., Magni, M., Cantoni, M., Piga, D., Bucca, G. (2020). Human–robot collaboration in sensorless assembly task learning enhanced by uncertainties adaptation via Bayesian optimization. Robotics and Autonomous Systems.
Human–robot collaboration in sensorless assembly task learning enhanced by uncertainties adaptation via Bayesian optimization
Authors: Roveda, L. and Magni, M. and Cantoni, M. and Piga, D. and Bucca, G.
Year: 2020
Abstract: Robots are increasingly exploited in production plants. Within the Industry 4.0 paradigm, the robot
complements the human’s capabilities, learning new tasks and adapting itself to compensate for
uncertainties. With this aim, the presented paper focuses on the investigation of machine learning
techniques to make a sensorless robot able to learn and optimize an industrial assembly task.
Relying on sensorless Cartesian impedance control, two main contributions are defined: (1) a task-
trajectory learning algorithm based on a few human’s demonstrations (exploiting Hidden Markov
Model approach), and (2) an autonomous optimization procedure of the task execution (exploiting
Bayesian Optimization). To validate the proposed methodology, an assembly task has been selected as
a reference application. The task consists of mounting a gear into its square-section shaft on a fixed
base to simulate the assembly of a gearbox. A Franka EMIKA Panda manipulator has been used as a
test platform, implementing the proposed methodology. The experiments, carried out on a population
of 15 subjects, show the effectiveness of the proposed strategy, making the robot able to learn and
optimize its behavior to accomplish the assembly task, even in the presence of task uncertainties.
Published in Robotics and Autonomous Systems.
Human–robot collaboration in sensorless assembly task learning enhanced by uncertainties adaptation via Bayesian optimization
@ARTICLE{Roveda2020j,
title = {Human--robot collaboration in sensorless assembly task learning enhanced by uncertainties adaptation via {B}ayesian optimization},
journal = {Robotics and Autonomous Systems},
author = {Roveda, L. and Magni, M. and Cantoni, M. and Piga, D. and Bucca, G.},
year = {2020},
doi = {10.1016/j.robot.2020.103711},
url = {}
}
DownloadRoveda, L., Magni, M., Cantoni, M., Piga, D., Bucca, G. (2020). Assembly task learning and optimization through Human’s demonstration and machine learning. In IEEE International Conference on Systems, Man, and Cybernetics.
Assembly task learning and optimization through Human’s demonstration and machine learning
Authors: Roveda, L. and Magni, M. and Cantoni, M. and Piga, D. and Bucca, G.
Year: 2020
Abstract: Robots are increasingly exploited in production plants, with the need to learn and to adapt themselves to new tasks. This paper focuses on the investigation of machine learning techniques to make a sensorless robot able to learn and optimize an industrial assembly task. Relying on sensorless Cartesian impedance control, a task-trajectory learning algorithm exploiting a limited number of human’s demonstrations (based on Hidden Markov Model), and an autonomous optimization procedure (based on Bayesian Optimization) are proposed to learn and optimize the assembly task. To validate the proposed methodology, an assembly task of a gear into its square-section shaft has been considered. A Franka EMIKA Panda manipulator has been used as a test platform. The experiments show the effectiveness of the proposed strategy, making the robot able to learn and optimize its behaviour to accomplish the assembly task, even in the presence of uncertainties.
Published in IEEE International Conference on Systems, Man, and Cybernetics.
Assembly task learning and optimization through Human’s demonstration and machine learning
@INPROCEEDINGS{Roveda2020k,
title = {Assembly task learning and optimization through {H}uman’s demonstration and machine learning},
booktitle = {{IEEE} International Conference on Systems, Man, and Cybernetics},
author = {Roveda, L. and Magni, M. and Cantoni, M. and Piga, D. and Bucca, G.},
year = {2020},
doi = {10.1109/SMC42975.2020.9282911},
url = {}
}
DownloadRoveda, L., Maskani, J., Franceschi, P., Arash, A., Braghin, F., Molinari Tosatti, L., Pedrocchi, N. (2020). Model-based reinforcement learning variable impedance control for human-robot collaboration. Journal of Intelligent & Robotic Systems.
Model-based reinforcement learning variable impedance control for human-robot collaboration
Authors: Roveda, L. and Maskani, J. and Franceschi, P. and Arash, A. and Braghin, F. and Molinari Tosatti, L. and Pedrocchi, N.
Year: 2020
Abstract: Industry 4.0 is taking human-robot collaboration at the center of the production environment. Collaborative robots
enhance productivity and flexibility while reducing human’s fatigue and the risk of injuries, exploiting advanced control
methodologies. However, there is a lack of real-time model-based controllers accounting for the complex human-robot
interaction dynamics. With this aim, this paper proposes a Model-Based Reinforcement Learning (MBRL) variable
impedance controller to assist human operators in collaborative tasks. More in details, an ensemble of Artificial Neural Networks (ANNs) is used to learn a human-robot interaction dynamic model, capturing uncertainties. Such a learned model is kept updated during collaborative tasks execution. In addition, the learned model is used by a Model Predictive Controller (MPC) with Cross-Entropy Method (CEM). The aim of the MPC+CEM is to online optimize the stiffness and damping impedance control parameters minimizing the human effort (i.e, minimizing the human-robot interaction forces). The proposed approach has been validated through an experimental procedure. A lifting task has been considered as the reference validation application (weight of the manipulated part: 10 kg unknown to the robot controller). A KUKA LBR iiwa 14 R820 has been used as a test platform. Qualitative performance (i.e, questionnaire on perceived collaboration) have been evaluated. Achieved results have been compared with previous developed offline model-free optimized controllers and with the robot manual guidance controller. The proposed MBRL variable impedance controller shows improved human-robot collaboration. The proposed controller is capable to actively assist the human in the target task, compensating for the unknown part weight. The human-robot interaction dynamic model has been trained with a few initial experiments (30 initial experiments). In addition, the possibility to keep the learning of the human-robot interaction dynamics active allows accounting for the adaptation of human motor system.
Published in Journal of Intelligent & Robotic Systems, Springer.
Model-based reinforcement learning variable impedance control for human-robot collaboration
@ARTICLE{Roveda2020c,
title = {Model-based reinforcement learning variable impedance control for human-robot collaboration},
journal = {Journal of Intelligent & Robotic Systems},
publisher = {Springer},
author = {Roveda, L. and Maskani, J. and Franceschi, P. and Arash, A. and Braghin, F. and Molinari Tosatti, L. and Pedrocchi, N.},
year = {2020},
doi = {10.1007/s10846- 020-01183-3},
url = {}
}
DownloadRoveda, L., Piga, D. (2020). Robust state dependent Riccati equation variable impedance control for robotic force-tracking tasks. International Journal of Intelligent Robotics and Applications.
Robust state dependent Riccati equation variable impedance control for robotic force-tracking tasks
Authors: Roveda, L. and Piga, D.
Year: 2020
Abstract: Industrial robots are increasingly used in highly flexible interaction tasks, where the intrinsic variability makes difficult to pre-program the manipulator for all the different scenarios. In such applications, interaction environments are commonly (partially) unknown to the robot, requiring the implemented controllers to take in charge for the stability of the interaction. While standard controllers are sensor-based, there is a growing need to make sensorless robots (i.e., most of the commercial robots are not equipped with force/torque sensors) able to sense the environment, properly reacting to the established interaction. This paper proposes a new methodology to sensorless force control manipulators. On the basis of sensorless Cartesian impedance control, an Extended Kalman Filter (EKF) is designed to estimate the interaction exchanged between the robot and the environment. Such an estimation is then used in order to close a robust high-performance force loop, designed exploiting a variable impedance control and a State Dependent Riccati Equation (SDRE) force controller. The described approach has been validated in simulations. A Franka EMIKA panda robot has been considered as a test platform. A probing task involving different materials (i.e., with different stiffness properties) has been considered to show the capabilities of the developed EKF (able to converge with limited errors) and controller (preserving stability and avoiding overshoots). The proposed controller has been compared with an LQR controller to show its improved performance.
Published in International Journal of Intelligent Robotics and Applications, Springer.
Robust state dependent Riccati equation variable impedance control for robotic force-tracking tasks
@ARTICLE{Roveda2020d,
title = {Robust state dependent {R}iccati equation variable impedance control for robotic force-tracking tasks},
journal = {International Journal of Intelligent Robotics and Applications},
publisher = {Springer},
author = {Roveda, L. and Piga, D.},
year = {2020},
doi = {10.1007/s41315-020-00153-0},
url = {}
}
DownloadRoveda, L., Piga, D. (2020). Interaction force computation exploiting environment stiffness estimation for sensorless robot applications. In IEEE Metrology for Industry 4.0 and IoT 2020.
Interaction force computation exploiting environment stiffness estimation for sensorless robot applications
Authors: Roveda, L. and Piga, D.
Year: 2020
Abstract: Industrial robots are increasingly used to perform tasks requiring an interaction with the surrounding environment.
However, standard controllers require force/torque measurements to close the loop. Most of the industrial manipulators do not have embedded force/torque sensor(s), requiring additional efforts (i.e., additional costs and implementation resources) for such integration in the robotic setup. To extend the use of compliant controllers to sensorless force control, a model-based methodology is presented in this paper. Relying on sensorless Cartesian impedance control, an Extended Kalman Filter (EKF) is proposed to estimate the interaction environment stiffness. Exploiting such estimation, the interaction force can be computed, e.g., to close the force loop, making the sensorless robot able to perform the target task (e.g., probing task, assembly task). The described approach has been validated with experiments. A Franka EMIKA panda robot has been used as a test platform. A probing task involving different materials (i.e., with different - unknown - stiffness properties) has been considered to show the capabilities of the developed EKF. The computed interaction force (on the basis of the estimated environment stiffness) has been compared with the Franka EMIKA panda robot force measurements to prove the effectiveness of the proposed approach.
Published in IEEE Metrology for Industry 4.0 and IoT 2020.
Interaction force computation exploiting environment stiffness estimation for sensorless robot applications
@INPROCEEDINGS{Roveda2020g,
title = {Interaction force computation exploiting environment stiffness estimation for sensorless robot applications},
booktitle = {{IEEE} Metrology for Industry 4.0 and {IoT} 2020},
author = {Roveda, L. and Piga, D.},
year = {2020},
doi = {10.1109/MetroInd4.0IoT48571.2020.9138189},
url = {}
}
DownloadRoveda, L., Savani, L., Arlati, S., Dinon, T., Legnani, G., Molinari Tosatti, L. (2020). Design methodology of an active back-support exoskeleton with adaptable backbone-based kinematics. International Journal of Industrial Ergonomics.
Design methodology of an active back-support exoskeleton with adaptable backbone-based kinematics
Authors: Roveda, L. and Savani, L. and Arlati, S. and Dinon, T. and Legnani, G. and Molinari Tosatti, L.
Year: 2020
Abstract: non-ergonomic conditions and to manipulate heavy parts. As a result, work-related musculoskeletal disorders are a major problem to tackle in workplace. In particular, back is one of the most affected regions. To solve such issue, many efforts have been made in the design and control of exoskeleton devices, relieving the human from the task load. Besides upper limbs and lower limbs exoskeletons, back-support exoskeletons have been also investigated, proposing both passive and active solutions. While passive solutions cannot empower the human’s capabilities, common active devices are rigid, without the possibility to track the human’s spine kinematics while executing the task. The here proposed paper describes a methodology to design an active back-support exoskeleton with backbone-based kinematics. On the basis of the (easily implementable) scissor hinge mechanism, a one-degree of freedom device has been designed. In particular, the resulting device allows tracking the motion of a reference vertebra, i.e., the vertebrae in the correspondence of the connection between the scissor hinge mechanism and the back of the operator. Therefore, the proposed device is capable to adapt to the human posture, guaranteeing the support while relieving the person from the task load. In addition, the proposed mechanism can be easily optimized and realized for different subjects, involving a subject-based design procedure, making possible to adapt its kinematics to track the spine motion of the specific user. A prototype of the proposed device has been 3D-printed to show the achieved kinematics. Preliminary tests for discomfort evaluation show the potential of the proposed methodology, foreseeing extensive subjects-based optimization, realization and testing of the device.
Published in International Journal of Industrial Ergonomics, Elsevier.
Design methodology of an active back-support exoskeleton with adaptable backbone-based kinematics
@ARTICLE{Roveda2020e,
title = {Design methodology of an active back-support exoskeleton with adaptable backbone-based kinematics},
journal = {International Journal of Industrial Ergonomics},
publisher = {Elsevier},
author = {Roveda, L. and Savani, L. and Arlati, S. and Dinon, T. and Legnani, G. and Molinari Tosatti, L.},
year = {2020},
doi = {10.1016/j.ergon.2020.102991},
url = {}
}
DownloadRuggieri, A., Stranieri, F., Stella, F., Scutari, M. (2020). Hard and soft em in bayesian network learning from incomplete data. Algorithms 13(12), 329.
Hard and soft em in bayesian network learning from incomplete data
Authors: Ruggieri, A. and Stranieri, F. and Stella, F. and Scutari, M.
Year: 2020
Published in Algorithms 13(12), 329.
Hard and soft em in bayesian network learning from incomplete data
@ARTICLE{scutari20h,
title = {Hard and soft em in bayesian network learning from incomplete data},
journal = {Algorithms},
volume = {13},
author = {Ruggieri, A. and Stranieri, F. and Stella, F. and Scutari, M.},
number = {12},
pages = {329},
year = {2020},
doi = {},
url = {}
}
DownloadSchürch, M., Azzimonti, D., Benavoli, A., Zaffalon, M. (2020). Recursive estimation for sparse gaussian process regression. Automatica 120, 109127.
Recursive estimation for sparse gaussian process regression
Authors: Schürch, M. and Azzimonti, D. and Benavoli, A. and Zaffalon, M.
Year: 2020
Abstract: Gaussian Processes (GPs) are powerful kernelized methods for non-parameteric regression used in many applications. However, their use is limited to a few thousand of training samples due to their cubic time complexity. In order to scale GPs to larger datasets, several sparse approximations based on so-called inducing points have been proposed in the literature. In this work we investigate the connection between a general class of sparse inducing point GP regression methods and Bayesian recursive estimation which enables Kalman Filter like updating for online learning. The majority of previous work has focused on the batch setting, in particular for learning the model parameters and the position of the inducing points, here instead we focus on training with mini-batches. By exploiting the Kalman filter formulation, we propose a novel approach that estimates such parameters by recursively propagating the analytical gradients of the posterior over mini-batches of the data. Compared to state of the art methods, our method keeps analytic updates for the mean and covariance of the posterior, thus reducing drastically the size of the optimization problem. We show that our method achieves faster convergence and superior performance compared to state of the art sequential Gaussian Process regression on synthetic GP as well as real-world data with up to a million of data samples.
Published in Automatica 120, Elsevier, 109127.
Recursive estimation for sparse gaussian process regression
@ARTICLE{schurch2020a,
title = {Recursive estimation for sparse gaussian process regression},
journal = {Automatica},
publisher = {Elsevier},
volume = {120},
author = {Sch\"urch, M. and Azzimonti, D. and Benavoli, A. and Zaffalon, M.},
pages = {109127},
year = {2020},
doi = {},
url = {}
}
DownloadShahid, A.A., Roveda, L., Piga, D., Braghin, F. (2020). Learning continuous control actions for robotic grasping with reinforcement learning. In IEEE International Conference on Systems, Man, and Cybernetics.
Learning continuous control actions for robotic grasping with reinforcement learning
Authors: Shahid, A.A. and Roveda, L. and Piga, D. and Braghin, F.
Year: 2020
Abstract: Robots are nowadays increasingly required to deal with (partially) unknown tasks and situations. The robot has, therefore, to adapt its behavior to the specific working conditions. Classical control methods in robotics require manually programming all actions of a robot. While very effective in fixed conditions, such model-based approaches cannot handle variations, demanding tedious tuning of parameters for every new task. Reinforcement learning (RL) holds the promise of autonomously learning new control policies through trial-and-error. However, RL approaches are prone to learning with high samples, particularly for continuous control problems. In this paper, a learning-based method is presented that leverages simulation data to learn an object manipulation task through RL. The control policy is parameterized by a neural network and learned using modern Proximal Policy Optimization (PPO) algorithm. A dense reward function has been designed for the task to enable efficient learning of an agent. The proposed approach is trained entirely in simulation (exploiting the MuJoCo environment) from scratch without any demonstrations of the task. A grasping task involving a Franka Emika Panda manipulator has been considered as the reference task to be learned. The task requires the robot to reach the part, grasp it, and lift it off the contact surface. The proposed approach has been demonstrated to be generalizable across multiple object geometries and initial robot/parts configurations, having the robot able to learn and re-execute the target task.
Published in IEEE International Conference on Systems, Man, and Cybernetics.
Learning continuous control actions for robotic grasping with reinforcement learning
@INPROCEEDINGS{Roveda2020l,
title = {Learning continuous control actions for robotic grasping with reinforcement learning},
booktitle = {{IEEE} International Conference on Systems, Man, and Cybernetics},
author = {Shahid, A.A. and Roveda, L. and Piga, D. and Braghin, F.},
year = {2020},
doi = {10.1109/SMC42975.2020.9282951},
url = {}
}
DownloadSheldrake, T.E., Caricchi, L., Scutari, M. (2020). Tectonic control on global variations in the record of large-magnitude explosive eruptions in volcanic arcs. Frontiers in Earth Sciences 8(127), pp. 1–14.
Tectonic control on global variations in the record of large-magnitude explosive eruptions in volcanic arcs
Authors: Sheldrake, T.E. and Caricchi, L. and Scutari, M.
Year: 2020
Published in Frontiers in Earth Sciences 8(127), pp. 1–14.
Tectonic control on global variations in the record of large-magnitude explosive eruptions in volcanic arcs
@ARTICLE{scutari20b,
title = {Tectonic control on global variations in the record of large-magnitude explosive eruptions in volcanic arcs},
journal = {Frontiers in Earth Sciences},
volume = {8},
author = {Sheldrake, T.E. and Caricchi, L. and Scutari, M.},
number = {127},
pages = {1--14},
year = {2020},
doi = {},
url = {}
}
DownloadSorgini, F., Airò Farulla, G., Lukic, N., Danilov, I., Roveda, L., Milivojevic, M., Babu Pulikottil, T., Carrozza, M.C., Prinetto, P., Tolio, T., Oddo, C.M., Petrovic, P., Bojovic, B. (2020). Tactile sensing with gesture-controlled collaborative robot. In 2020 IEEE International Workshop on Metrology for Industry 4.0 & IoT (MetroInd4.0 & IoT).
Tactile sensing with gesture-controlled collaborative robot
Authors: Sorgini, F. and Airò Farulla, G. and Lukic, N. and Danilov, I. and Roveda, L. and Milivojevic, M. and Babu Pulikottil, T. and Carrozza, M.C. and Prinetto, P. and Tolio, T. and Oddo, C.M. and Petrovic, P. and Bojovic, B.
Year: 2020
Abstract: Sensors and human machine interfaces for
collaborative robotics will allow smooth interaction in contexts ranging from industry to tele-medicine and rescue. This paper introduces a bidirectional communication system to achieve multisensory telepresence during the gestural control of an industrial robotic arm. Force and motion from the robot are converted in neuromorphic haptic stimuli delivered on the user’s hand through a vibro-tactile glove. Untrained personnel participated in an experimental task benchmarking a pick-and-place operation. The robot end-effector was used to
sequentially press six buttons, illuminated according to a
random sequence, and comparing the tasks executed without
and with tactile feedback. The results demonstrated the
reliability of the hand tracking strategy developed for
controlling the robotic arm, and the effectiveness of a neuronal spiking model for encoding hand displacement and exerted forces in order to promote a fluid embodiment of the haptic interface and control strategy. The main contribution of this paper is in presenting a robotic arm under gesture-based remote control with multisensory telepresence, demonstrating for the first time that a spiking haptic interface can be used to effectively deliver on the skin surface a sequence of stimuli emulating the neural code of the mechanoreceptors beneath.
Published in 2020 IEEE International Workshop on Metrology for Industry 4.0 & IoT (MetroInd4.0 & IoT).
Tactile sensing with gesture-controlled collaborative robot
@INPROCEEDINGS{Roveda2020i,
title = {Tactile sensing with gesture-controlled collaborative robot},
booktitle = {2020 {IEEE} International Workshop on Metrology for Industry 4.0 & {IoT} ({MetroInd4}.0 & {IoT})},
author = {Sorgini, F. and Air\`o Farulla, G. and Lukic, N. and Danilov, I. and Roveda, L. and Milivojevic, M. and Babu Pulikottil, T. and Carrozza, M.C. and Prinetto, P. and Tolio, T. and Oddo, C.M. and Petrovic, P. and Bojovic, B.},
year = {2020},
doi = {10.1109/MetroInd4.0IoT48571.2020.9138183},
url = {}
}
DownloadVolpetti, C., Kanjirangat, V., Antonucci, A. (2020). Temporal word embeddings for narrative understanding. In 12th International Conference on Machine Learning and Computing (ICMLC 2020), ACM.
Temporal word embeddings for narrative understanding
Authors: Volpetti, C. and Kanjirangat, V. and Antonucci, A.
Year: 2020
Abstract: We propose temporal word embeddings as a suitable tool to study the evolution of characters and their sentiments across the plot of a narrative text. The dynamic evolution of instances within a narrative text is a challenging task, where complex behavioral evolutions and other characteristics specific to the narrative text
need to be inferred and interpreted. While starting from an existing approach to the learning of these models, we propose an alternative initialization procedure which seems to be especially suited for the case of narrative text. As a validation benchmark, we use the Harry Potter series of books as a challenging case study for such character
trait evolution. A benchmark data set based on temporal word
analogies related to the characters in the plot of the series is considered. The results are promising, and the empirical validation seems to support the working ideas behind this proposal.
Accepted in 12th International Conference on Machine Learning and Computing (ICMLC 2020), ACM.
Temporal word embeddings for narrative understanding
@INPROCEEDINGS{supsi2020a,
title = {Temporal word embeddings for narrative understanding},
publisher = {ACM},
booktitle = {12th International Conference on Machine Learning and Computing ({ICMLC} 2020)},
author = {Volpetti, C. and Kanjirangat, V. and Antonucci, A.},
year = {2020},
doi = {},
url = {http://www.icmlc.org/}
}
DownloadYoo, J., Kang, U., Scanagatta, M., Corani, G., Zaffalon, M. (2020). Sampling subgraphs with guaranteed treewidth for accurate and efficient graphical inference. In Proceedings of International Conference on Web Search and Data Mining (WSDM '20).
Sampling subgraphs with guaranteed treewidth for accurate and efficient graphical inference
Authors: Yoo, J. and Kang, U. and Scanagatta, M. and Corani, G. and Zaffalon, M.
Year: 2020
Abstract: How can we run graphical inference on large graphs efficiently and accurately? Many real-world networks are modeled as graphical models, and graphical inference is fundamental to understand the properties of those networks. In this work, we propose a novel approach for fast and accurate inference, which first samples a small subgraph and then runs inference over the subgraph instead of the given graph. This is done by the bounded treewidth (BTW) sampling, our novel algorithm that generates a subgraph with guar- anteed bounded treewidth while retaining as many edges as pos- sible. We first analyze the properties of BTW theoretically. Then, we evaluate our approach on node classification and compare it with the baseline which is to run loopy belief propagation (LBP) on the original graph. Our approach can be coupled with various inference algorithms: it shows higher accuracy up to 13.7% with the junction tree algorithm, and allows faster inference up to 23.8 times with LBP. We further compare BTW with previous graph sampling algorithms and show that it gives the best accuracy.
Accepted in Proceedings of International Conference on Web Search and Data Mining (WSDM '20).
Sampling subgraphs with guaranteed treewidth for accurate and efficient graphical inference
@INPROCEEDINGS{corani2019e,
title = {Sampling subgraphs with guaranteed treewidth for accurate and efficient graphical inference},
booktitle = {Proceedings of International Conference on Web Search and Data Mining ({WSDM} '20)},
author = {Yoo, J. and Kang, U. and Scanagatta, M. and Corani, G. and Zaffalon, M.},
year = {2020},
doi = {},
url = {https://dl.acm.org/doi/pdf/10.1145/3336191.3371815}
}
DownloadZaffalon, M., Antonucci, A., Cabañas, R. (2020). Structural causal models are (solvable by) credal networks. In Proceedings of the 10th International Conference on Probabilistic Graphical Models (PGM 2020), JMLR.org.
Structural causal models are (solvable by) credal networks
Authors: Zaffalon, M. and Antonucci, A. and Cabañas, R.
Year: 2020
Abstract: A structural causal model is made of endogenous (manifest) and exogenous (latent) variables. We show that endogenous observations induce linear constraints on the probabilities of the exogenous variables. This allows to exactly map a causal model into a credal network. Causal inferences, such as interventions and counterfactuals, can consequently be obtained by standard algorithms for the updating of credal nets. These natively return sharp values in the identifiable case, while intervals corresponding to the exact bounds are produced for unidentifiable queries. A characterization of the causal models that allow the map above to be compactly derived is given, along with a discussion about the scalability for general models. This contribution should be regarded as a systematic approach to represent structural causal models by credal networks and hence to systematically compute causal inferences. A number of demonstrative examples is presented to clarify our methodology. Extensive experiments show that approximate algorithms for credal networks can immediately be used to do causal inference in real-size problems.
Accepted in Proceedings of the 10th International Conference on Probabilistic Graphical Models (PGM 2020), JMLR.org.
Structural causal models are (solvable by) credal networks
@INPROCEEDINGS{zaffalon2020b,
title = {Structural causal models are (solvable by) credal networks},
publisher = {JMLR.org},
booktitle = {Proceedings of the 10th International Conference on Probabilistic Graphical Models ({PGM} 2020)},
author = {Zaffalon, M. and Antonucci, A. and Cabañas, R.},
year = {2020},
doi = {},
url = {https://pgm2020.cs.aau.dk/}
}
Download top2019
Antonucci, A. (2019). Reliable discretisation of deterministic equations in Bayesian networks. In Proceedings of the 32nd International Flairs Conference (FLAIRS-32), AAAI Press.
Reliable discretisation of deterministic equations in Bayesian networks
Authors: Antonucci, A.
Year: 2019
Abstract: We focus on the problem of modeling deterministic equations over continuous variables in discrete Bayesian networks. This is typically achieved by a discretisation of both input and output variables and a degenerate quantification of the corre- sponding conditional probability tables. This approach, based on classical probabilities, cannot properly model the information loss induced by the discretisation. We show that a reli- able modeling of such epistemic uncertainty can be instead achieved by credal sets, i.e., convex sets of probability mass functions. This transforms the original Bayesian network in a credal network, possibly returning interval-valued inferences, that are robust with respect to the information loss induced by the discretisation. Algorithmic strategies for an optimal choice of the discretisation bins are also discussed.
Published in Proceedings of the 32nd International Flairs Conference (FLAIRS-32), AAAI Press.
Note: Accepted for pubblication.
Reliable discretisation of deterministic equations in Bayesian networks
@INPROCEEDINGS{supsi2019c,
title = {Reliable discretisation of deterministic equations in {B}ayesian networks},
publisher = {AAAI Press},
booktitle = {Proceedings of the 32nd International Flairs Conference ({FLAIRS}-32)},
author = {Antonucci, A.},
year = {2019},
doi = {},
url = {https://aaai.org/ocs/index.php/FLAIRS/FLAIRS19/paper/view/18228/17346}
}
DownloadAntonucci, A., Facchini, A., Mattei, L. (2019). Credal sentential decision diagrams. In Proceedings of the Eleventh International Symposium on Imprecise Probability: Theories and Applications (ISIPTA '19).
Credal sentential decision diagrams
Authors: Antonucci, A. and Facchini, A. and Mattei, L.
Year: 2019
Abstract: Probabilistic sentential decision diagrams are logical circuits annotated by probability mass functions on the disjunctive gates. This allows for a compact representation of joint mass functions consistent with logical constraints. We propose a credal generalisation of the probabilistic quantification of these models, that allows to replace the local probabilities with (credal) sets of mass functions specified by linear constraints. This induces a joint credal set, that sharply assigns probability zero to states inconsistent with the constraints. These models can support cautious estimates of the local parameters when only small amounts of training data are available. Algorithmic strategies to compute lower and upper bounds of marginal and conditional queries are provided. The task can be achieved in lin- ear time with respect to the diagram size for marginal queries. The same can be done for conditional queries if the topology of the circuit is singly connected.
Published in Proceedings of the Eleventh International Symposium on Imprecise Probability: Theories and Applications (ISIPTA '19).
Note: Accepted for pubblication.
Credal sentential decision diagrams
@INPROCEEDINGS{supsi2019b,
title = {Credal sentential decision diagrams},
booktitle = {Proceedings of the Eleventh International Symposium on Imprecise Probability: Theories and Applications ({ISIPTA} '19)},
author = {Antonucci, A. and Facchini, A. and Mattei, L.},
year = {2019},
doi = {},
url = {http://www.isipta2019.ugent.be/contributions/antonucci19.pdf}
}
DownloadArnone, E., Azzimonti, L., Nobile, F., Sangalli, L.M. (2019). Modeling spatially dependent functional data via regression with differential regularization. Journal of Multivariate Analysis 170, pp. 275-295.
Modeling spatially dependent functional data via regression with differential regularization
Authors: Arnone, E. and Azzimonti, L. and Nobile, F. and Sangalli, L.M.
Year: 2019
Abstract: We propose a method for modeling spatially dependent functional data, based on regression with differential regularization. The regularizing term enables to include problem-specific information about the spatio-temporal variation of the phenomenon under study, formalized in terms of a time-dependent partial differential equation. The method is implemented using a discretization based on finite elements in space and finite differences in time. This non-tensor product basis allows to handle efficiently data distributed over complex domains and where the shape of the domain influences the phenomenon’s behavior. Moreover, the method can comply with specific conditions at the boundary of the domain of interest. Simulation studies compare the proposed model to available techniques for spatio-temporal data. The method is also illustrated via an application to the study of blood-flow velocity field in a carotid artery affected by atherosclerosis, starting from echo-color doppler and magnetic resonance imaging data.
Published in Journal of Multivariate Analysis 170, pp. 275-295.
Note: Special Issue on Functional Data Analysis and Related Topics
Modeling spatially dependent functional data via regression with differential regularization
@ARTICLE{azzimonti2018a,
title = {Modeling spatially dependent functional data via regression with differential regularization},
journal = {Journal of Multivariate Analysis},
volume = {170},
author = {Arnone, E. and Azzimonti, L. and Nobile, F. and Sangalli, L.M.},
pages = {275-295},
year = {2019},
doi = {10.1016/j.jmva.2018.09.006},
url = {}
}
DownloadAzzimonti, L., Corani, G., Zaffalon, M. (2019). Hierarchical estimation of parameters in Bayesian networks. Computational Statistics and Data Analysis 137, pp. 67–91.
Hierarchical estimation of parameters in Bayesian networks
Authors: Azzimonti, L. and Corani, G. and Zaffalon, M.
Year: 2019
Abstract: A novel approach for parameter estimation in Bayesian networks is presented. The main idea is to introduce a hyper-prior in the Multinomial-Dirichlet model, traditionally used for conditional distribution estimation in Bayesian networks. The resulting hierarchical model jointly estimates different conditional distributions belonging to the same conditional probability table, thus borrowing statistical strength from each other. An analytical study of the dependence structure a priori induced by the hierarchical model is performed and an ad hoc variational algorithm for fast and accurate inference is derived. The proposed hierarchical model yields a major performance improvement in classification with Bayesian networks compared to traditional models. The proposed variational algorithm reduces by two orders of magnitude the computational time, with the same accuracy in parameter estimation, compared to traditional MCMC methods.
Moreover, motivated by a real case study, the hierarchical model is applied to the estimation of Bayesian networks parameters by borrowing strength from related domains.
Published in Computational Statistics and Data Analysis 137, pp. 67–91.
Hierarchical estimation of parameters in Bayesian networks
@ARTICLE{azzimonti2019a,
title = {Hierarchical estimation of parameters in {B}ayesian networks},
journal = {Computational Statistics and Data Analysis},
volume = {137},
author = {Azzimonti, L. and Corani, G. and Zaffalon, M.},
pages = {67--91},
year = {2019},
doi = {10.1016/j.csda.2019.02.004},
url = {}
}
DownloadAzzimonti, D., Ginsbourger, D., Chevalier, C., Bect, J., Richet, Y. (2019). Adaptive design of experiments for conservative estimation of excursion sets. Technometrics 63(1), pp. 13-26.
Adaptive design of experiments for conservative estimation of excursion sets
Authors: Azzimonti, D. and Ginsbourger, D. and Chevalier, C. and Bect, J. and Richet, Y.
Year: 2019
Abstract: We consider the problem of estimating the set of all inputs that leads a system to some particular behavior. The system is modeled by an expensive-to-evaluate function, such as a computer experiment, and we are interested in its excursion set, i.e. the set of points where the function takes values above or below some prescribed threshold. The objective function is emulated with a Gaussian Process (GP) model based on an initial design of experiments enriched with evaluation results at (batch-) sequentially determined input points. The GP model provides conservative estimates for the excursion set, which control false positives while minimizing false negatives. We introduce adaptive strategies that sequentially select new evaluations of the function by reducing the uncertainty on conservative estimates. Following the Stepwise Uncertainty Reduction approach we obtain new evaluations by minimizing adapted criteria. Tractable formulae for the conservative criteria are derived, which allow more convenient optimization. The method is benchmarked on random functions generated under the model assumptions in different scenarios of noise and batch size. We then apply it to a reliability engineering test case. Overall, the proposed strategy of minimizing false negatives in conservative estimation achieves competitive performance both in terms of model-based and model-free indicators.
Published in Technometrics 63(1), Taylor & Francis, pp. 13-26.
Adaptive design of experiments for conservative estimation of excursion sets
@ARTICLE{azzimontid2019c,
title = {Adaptive design of experiments for conservative estimation of excursion sets},
journal = {Technometrics},
publisher = {Taylor & Francis},
volume = {63},
author = {Azzimonti, D. and Ginsbourger, D. and Chevalier, C. and Bect, J. and Richet, Y.},
number = {1},
pages = {13-26},
year = {2019},
doi = {10.1080/00401706.2019.1693427},
url = {}
}
DownloadAzzimonti, D., Ginsbourger, D., Rohmer, J., Idier, D. (2019). Profile extrema for visualizing and quantifying uncertainties on excursion regions. Application to coastal flooding. Technometrics 61(4), pp. 474–493.
Profile extrema for visualizing and quantifying uncertainties on excursion regions. Application to coastal flooding
Authors: Azzimonti, D. and Ginsbourger, D. and Rohmer, J. and Idier, D.
Year: 2019
Abstract: We consider the problem of describing excursion sets of a real-valued function f, i.e. the set of inputs where f is above a fixed threshold. Such regions are hard to visualize if the input space dimension, d, is higher than 2. For a given projection matrix from the input space to a lower dimensional (usually 1, 2) subspace, we introduce profile sup (inf) functions that associate to each point in the projection’s image the sup (inf) of the function constrained over the pre-image of this point by the considered projection. Plots of profile extrema functions convey a simple, although intrinsically partial, visualization of the set. We consider expensive to evaluate functions where only a very limited number of evaluations, n, is available, e.g. n<100d, and we surrogate f with a posterior quantity of a Gaussian process (GP) model. We first compute profile extrema functions for the posterior mean given n evaluations of f. We quantify the uncertainty on such estimates by studying the distribution of GP profile extrema with posterior quasi-realizations obtained from an approximating process. We control such approximation with a bound inherited from the Borell-TIS inequality. The technique is applied to analytical functions (d = 2, 3) and to a 5-dimensional coastal flooding test case for a site located on the Atlantic French coast. Here f is a numerical model returning the area of flooded surface in the coastal region given some offshore conditions. Profile extrema functions allowed us to better understand which offshore conditions impact large flooding events.
Published in Technometrics 61(4), Taylor & Francis, pp. 474–493.
Profile extrema for visualizing and quantifying uncertainties on excursion regions. Application to coastal flooding
@ARTICLE{azzimontid2019a,
title = {Profile extrema for visualizing and quantifying uncertainties on excursion regions. Application to coastal flooding},
journal = {Technometrics},
publisher = {Taylor & Francis},
volume = {61},
author = {Azzimonti, D. and Ginsbourger, D. and Rohmer, J. and Idier, D.},
number = {4},
pages = {474--493},
year = {2019},
doi = {10.1080/00401706.2018.1562987},
url = {}
}
DownloadAzzimonti, D., Rottondi, C., Tornatore, M. (2019). Using active learning to decrease probes for QoT estimation in optical networks. In , Optical Society of America, Th1H.1.
Using active learning to decrease probes for QoT estimation in optical networks
Authors: Azzimonti, D. and Rottondi, C. and Tornatore, M.
Year: 2019
Abstract: We use active learning to reduce the number of probes needed for machine learning-based QoT estimation. When building an estimation model based on Gaussian processes, only QoT instances that minimize estimation uncertainty are iteratively requested.
Published in Optical Fiber Communication Conference (OFC) 2019, Optical Society of America, Th1H.1.
Using active learning to decrease probes for QoT estimation in optical networks
@INPROCEEDINGS{azzimontid2019b,
title = {Using active learning to decrease probes for {QoT} estimation in optical networks},
journal = {Optical Fiber Communication Conference ({OFC}) 2019},
publisher = {Optical Society of America},
author = {Azzimonti, D. and Rottondi, C. and Tornatore, M.},
pages = {Th1H.1},
year = {2019},
doi = {10.1364/OFC.2019.Th1H.1},
url = {}
}
DownloadBenavoli, A., Facchini, A., Zaffalon, M. (2019). Bernstein's socks, polynomial-time provable coherence and entanglement. In De Bock, J., de Campos, C., de Cooman, G., Quaeghebeur, E., Wheeler, G. (Eds), ISIPTA '19: Proceedings of the Eleventh International Symposium on Imprecise Probability: Theories and Applications, PMLR 103, JMLR.org, pp. 23-31.
Bernstein's socks, polynomial-time provable coherence and entanglement
Authors: Benavoli, A. and Facchini, A. and Zaffalon, M.
Year: 2019
Abstract: We recently introduced a bounded rationality approach for the theory of desirable gambles. It is based on the unique requirement that being nonnegative for a gamble has to be defined so that it can be provable in polynomial time. In this paper we continue to investigate properties of this class of models. In particular we verify that the space of Bernstein polynomials in which nonnegativity is specified by the Krivine-Vasilescu certificate is yet another instance of this theory. As a consequence, we show how it is possible to construct in it a thought experiment uncovering entanglement with classical (hence non quantum) coins.
Accepted in De Bock, J., de Campos, C., de Cooman, G., Quaeghebeur, E., Wheeler, G. (Eds), ISIPTA '19: Proceedings of the Eleventh International Symposium on Imprecise Probability: Theories and Applications, PMLR 103, JMLR.org, pp. 23-31.
Bernstein's socks, polynomial-time provable coherence and entanglement
@INPROCEEDINGS{zaffalon2019b,
title = {Bernstein's socks, polynomial-time provable coherence and entanglement},
editor = {De Bock, J. and de Campos, C. and de Cooman, G. and Quaeghebeur, E. and Wheeler, G.},
publisher = {JMLR.org},
series = {PMLR},
volume = {103},
booktitle = {{ISIPTA };'19: Proceedings of the Eleventh International Symposium on Imprecise Probability: Theories and Applications},
author = {Benavoli, A. and Facchini, A. and Zaffalon, M.},
pages = {23-31},
year = {2019},
doi = {},
url = {http://proceedings.mlr.press/v103/benavoli19a/benavoli19a.pdf}
}
DownloadBolt, J.H., van der Gaag, L.C. (2019). On minimum elementary-triplet bases for independence relations. In De Bock, J., de Campos, C.P., de Cooman, G., Quaeghebeur, E., Wheeler, G. (Eds), Proceedings of the Eleventh International Symposium on Imprecise Probability: Theories and Applications (ISIPTA '19), PMLR 103, JMLR.org, pp. 32–37.
On minimum elementary-triplet bases for independence relations
Authors: Bolt, J.H. and van der Gaag, L.C.
Year: 2019
Abstract: A semi-graphoid independence relation is a set of independence statements, called triplets, and is typically exponentially large in the number of variables involved. For concise representation of such a relation, a subset of its triplets is listed in a so-called basis; its other triplets are defined implicitly through a set of axioms. An elementary-triplet basis for this purpose consists of all elementary triplets of a relation. Such a basis however, may include redundant information. In this paper we provide two lower bounds on the size of an elementary-triplet basis in general and an upper bound on the size of a minimum elementary-triplet basis. We further specify the construction of an elementary-triplet basis of minimum size for restricted relations.
Keywords: Independence relations, Axioms of independence, Elementary triplets, Basis representation.
Published in De Bock, J., de Campos, C.P., de Cooman, G., Quaeghebeur, E., Wheeler, G. (Eds), Proceedings of the Eleventh International Symposium on Imprecise Probability: Theories and Applications (ISIPTA '19), PMLR 103, JMLR.org, pp. 32–37.
On minimum elementary-triplet bases for independence relations
@INPROCEEDINGS{Linda2019c,
title = {On minimum elementary-triplet bases for independence relations},
editor = {De Bock, J. and de Campos, C.P. and de Cooman, G. and Quaeghebeur, E. and Wheeler, G.},
publisher = {JMLR.org},
series = {PMLR},
volume = {103},
booktitle = {Proceedings of the Eleventh International Symposium on Imprecise Probability: Theories and Applications ({ISIPTA} '19)},
author = {Bolt, J.H. and van der Gaag, L.C.},
pages = {32--37},
year = {2019},
doi = {},
url = {http://www.sipta.org/isipta19/contributions/bolt19.pdf}
}
DownloadBreschi, V., Piga, D., Bemporad, A. (2019). Online end-use energy disaggregation via jump linear models. Control Engineering Practice 89, pp. 30–42.
Online end-use energy disaggregation via jump linear models
Authors: Breschi, V. and Piga, D. and Bemporad, A.
Year: 2019
Abstract: This paper presents two iterative algorithms for non-intrusive appliance load monitoring, which aims to decompose the aggregate power consumption only measured at the household level into the contributions of the individual electric appliances. The approaches are based on modelling the total power consumption as a combination of jump linear sub-models, each of them describing the behaviour of the individual appliance. Dynamic-programming and multi-model Kalman filtering techniques are used to reconstruct the power consumptions at the single-appliance level from the aggregate power in an iterative way.
Published in Control Engineering Practice 89, pp. 30–42.
Online end-use energy disaggregation via jump linear models
@ARTICLE{piga2019b,
title = {Online end-use energy disaggregation via jump linear models},
journal = {Control Engineering Practice},
volume = {89},
author = {Breschi, V. and Piga, D. and Bemporad, A.},
pages = {30--42},
year = {2019},
doi = {10.1016/j.conengprac.2019.05.011},
url = {}
}
DownloadBucher, D., Mangili, F., Cellina, F., Bonesana, C., Jonietz, D., Raubal, M. (2019). From location tracking to personalized eco-feedback: a framework for geographic information collection, processing and visualization to promote sustainable mobility behaviors. Travel Behaviour and Society 14, pp. 43-56.
From location tracking to personalized eco-feedback: a framework for geographic information collection, processing and visualization to promote sustainable mobility behaviors
Authors: Bucher, D. and Mangili, F. and Cellina, F. and Bonesana, C. and Jonietz, D. and Raubal, M.
Year: 2019
Abstract: Nowadays, most people carry around a powerful smartphone which is well suited to constantly monitor the location and sometimes even the activity of its user. This makes tracking prevalent and leads to a large number of projects concerned with trajectory data. One area of particular interest is transport and mobility, where data is important for urban planning and smart city-related activities, but can also be used to provide individual users with feedback and suggestions for personal behavior change. As part of a large-scale study based in Switzerland, we use activity tracking data to provide people with eco-feedback on their own mobility patterns and stimulate them to adopt more energy-efficient mobility choices. In this paper we explore the opportunities offered by smartphone based activity tracking, propose a general framework to exploit location data to foster more sustainable mobility behavior, describe the technical solutions chosen and discuss a range of outcomes in terms of user perception and sustainability potential. The presented approach extracts mobility patterns from users’ trajectories, computes credible alternative transport options, and presents the results in a concise and clear way. The resulting eco-feedback helps people to understand their mobility choices, discover the most non-ecological parts of their travel behavior, and explore feasible alternatives.
Published in Travel Behaviour and Society 14, pp. 43-56.
From location tracking to personalized eco-feedback: a framework for geographic information collection, processing and visualization to promote sustainable mobility behaviors
@ARTICLE{mangili2019a,
title = { From location tracking to personalized eco-feedback: a framework for geographic information collection, processing and visualization to promote sustainable mobility behaviors},
journal = {Travel Behaviour and Society},
volume = {14},
author = {Bucher, D. and Mangili, F. and Cellina, F. and Bonesana, C. and Jonietz, D. and Raubal, M.},
pages = {43-56},
year = {2019},
doi = {10.1016/j.tbs.2018.09.005},
url = {}
}
DownloadCarollo, V., Piga, D., Borri, C., Paggi, M. (2019). Identification of elasto-plastic and nonlinear fracture mechanics parameters of silver-plated copper busbars for photovoltaics. Engineering Fracture Mechanics 205, pp. 439–454.
Identification of elasto-plastic and nonlinear fracture mechanics parameters of silver-plated copper busbars for photovoltaics
Authors: Carollo, V. and Piga, D. and Borri, C. and Paggi, M.
Year: 2019
Abstract: Silver-plated copper busbars are screen printed onto silicon solar cells and have the key role to collect the electric current produced by the solar cell. Busbars of two adjacent solar cells are then connected by a soldered ribbon made of the same material. Due to mechanical and thermal loads, such a ribbon is subject to axial deformation that, often, causes plasticity and, in some cases, its breakage due to crack growth. A procedure based on the gradient-descent method and particle swarm optimization is herein proposed for the identification of elasto-plastic and nonlinear (cohesive zone model, CZM) fracture mechanics parameters of silver-plated copper busbars. The proposed method requires the experimental determination of the force-displacement curves from uniaxial tensile tests on busbar samples with and without initial notches. The inspection of in situ SEM images during the tests allows also the estimation of the crack opening, which is found to be an important local quantity to assess the reliability of different CZMs in simulating a crack growth process consistent with the real one.
Published in Engineering Fracture Mechanics 205, pp. 439–454.
Identification of elasto-plastic and nonlinear fracture mechanics parameters of silver-plated copper busbars for photovoltaics
@ARTICLE{piga2019d,
title = {Identification of elasto-plastic and nonlinear fracture mechanics parameters of silver-plated copper busbars for photovoltaics},
journal = {Engineering Fracture Mechanics},
volume = {205},
author = {Carollo, V. and Piga, D. and Borri, C. and Paggi, M.},
pages = {439--454},
year = {2019},
doi = {},
url = {https://www.sciencedirect.com/science/article/pii/S001379441830451X?via%3Dihub}
}
DownloadCellina, F., Bucher, D., Mangili, F., Simão, J.V., Rudel, R., Raubal, M. (2019). A large scale, app-based behaviour change experiment persuading sustainable mobility patterns: methods, results and lessons learnt. Sustainability 11(9), 2674.
A large scale, app-based behaviour change experiment persuading sustainable mobility patterns: methods, results and lessons learnt
Authors: Cellina, F. and Bucher, D. and Mangili, F. and Simão, J.V. and Rudel, R. and Raubal, M.
Year: 2019
Abstract: The present urban transportation system, mostly tailored for cars, has long shown its limitations. In many urban areas, public transportation and soft mobility would be able to effectively satisfy many travel needs. However, they tend to be neglected, due to a deep-rooted car dependency. How can we encourage people to make sustainable mobility choices, reducing car use and the related CO 2 emissions and energy consumption? Taking advantage of the wide availability of smartphone devices, we designed GoEco!, a smartphone application exploiting automatic mobility tracking, eco-feedback, social comparison and gamification elements to persuade individual modal change. We tested the effectiveness of GoEco! in two regions of Switzerland (Cantons Ticino and Zurich), in a large-scale, one year long randomized controlled trial. Notwithstanding a large drop-out rate experienced throughout the experiment, GoEco! was observed to produce a statistically significant impact (a decrease in CO 2 emissions and energy consumption per kilometer) for systematic routes in highly car-dependent urban areas, such as the Canton Ticino. In Zurich, instead, where high quality public transport is already available, no statistically significant effects were found. In this paper we present the GoEco! experiment and discuss its results and the lessons learnt, highlighting practical difficulties in performing randomized controlled trials in the field of mobility and providing recommendations for future research.
Published in mdpi (Ed), Sustainability 11(9), 2674.
Note: Special Issue Environmental and Behavioral Consequences of Interventions for Sustainable Travel
A large scale, app-based behaviour change experiment persuading sustainable mobility patterns: methods, results and lessons learnt
@ARTICLE{mangili2019b,
title = {A large scale, app-based behaviour change experiment persuading sustainable mobility patterns: methods, results and lessons learnt},
journal = {Sustainability},
editor = {mdpi},
volume = {11},
author = {Cellina, F. and Bucher, D. and Mangili, F. and Sim\~ao, J.V. and Rudel, R. and Raubal, M.},
number = {9},
pages = {2674},
year = {2019},
doi = {https://doi.org/10.3390/su11092674},
url = {}
}
DownloadColic, N., Rinaldi, F. (2019). Improving spaCy dependency annotation and PoS tagging web service using independent NER services. Genomics Inform 17(2), e21.
Improving spaCy dependency annotation and PoS tagging web service using independent NER services
Authors: Colic, N. and Rinaldi, F.
Year: 2019
Abstract: Dependency parsing is often used as a component in many text analysis pipelines. However, performance, especially in specialized domains, suffers from the presence of complex terminology. Our hypothesis is that including named entity annotations can improve the speed and quality of dependency parses. As part of BLAH5, we built a web service delivering improved dependency parses by taking into account named entity annotations obtained by third party services. Our evaluation shows improved results and better speed.
Published in Genomics Inform 17(2), e21.
Improving spaCy dependency annotation and PoS tagging web service using independent NER services
@ARTICLE{rinaldi2019c,
title = {Improving {spaCy} dependency annotation and {PoS} tagging web service using independent {NER} services},
journal = {Genomics Inform},
volume = {17},
author = {Colic, N. and Rinaldi, F.},
number = {2},
pages = {e21},
year = {2019},
doi = {10.5808/GI.2019.17.2.e21},
url = {}
}
DownloadCorreia, A.H.C., de Campos, C.P., van der Gaag, L.C. (2019). An experimental study of prior dependence in Bayesian network structure learning. In De Bock, J., de Campos, C.P., de Cooman, G., Quaeghebeur, E., Wheeler, G. (Eds), Proceedings of the Eleventh International Symposium on Imprecise Probability: Theories and Applications (ISIPTA '19), PMLR 103, JMLR.org, pp. 78–81.
An experimental study of prior dependence in Bayesian network structure learning
Authors: Correia, A.H.C. and de Campos, C.P. and van der Gaag, L.C.
Year: 2019
Abstract: The Bayesian Dirichlet equivalent uniform (BDeu) function is a popular score to evaluate the goodness of a Bayesian network structure given complete categorical data. Despite its interesting properties, such as likelihood equivalence, it does require a prior expressed via a user-defined parameter known as Equivalent Sample Size (ESS), which significantly affects the final structure. We study conditions to obtain prior independence in BDeu-based structure learning. We show in experiments that the amount of data needed to render the learning robust to different ESS values is prohibitively large, even in big data times.
Keywords: Robustness, Bayesian Networks, Structure Learning, BDeu
Published in De Bock, J., de Campos, C.P., de Cooman, G., Quaeghebeur, E., Wheeler, G. (Eds), Proceedings of the Eleventh International Symposium on Imprecise Probability: Theories and Applications (ISIPTA '19), PMLR 103, JMLR.org, pp. 78–81.
An experimental study of prior dependence in Bayesian network structure learning
@INPROCEEDINGS{Linda2910d,
title = {An experimental study of prior dependence in {B}ayesian network structure learning},
editor = {De Bock, J. and de Campos, C.P. and de Cooman, G. and Quaeghebeur, E. and Wheeler, G.},
publisher = {JMLR.org},
series = {PMLR},
volume = {103},
booktitle = {Proceedings of the Eleventh International Symposium on Imprecise Probability: Theories and Applications ({ISIPTA} '19)},
author = {Correia, A.H.C. and de Campos, C.P. and van der Gaag, L.C.},
pages = {78--81},
year = {2019},
doi = {},
url = {http://www.sipta.org/isipta19/contributions/correia19.pdf}
}
DownloadEllendorff, T., Furrer, L., Colic, N., Aepli, N., Rinaldi, F. (2019). Approaching SMM4H with merged models and multi-task learning. In Proceedings of the Fourth Social Media Mining for Health Applications (#smm4h) Workshop & Shared Task, Association for Computational Linguistics, pp. 58–61.
Approaching SMM4H with merged models and multi-task learning
Authors: Ellendorff, T. and Furrer, L. and Colic, N. and Aepli, N. and Rinaldi, F.
Year: 2019
Abstract: We describe our submissions to the 4th edi- tion of the Social Media Mining for Health Applications (SMM4H) shared task. Our team (UZH) participated in two sub-tasks: Automatic classifications of adverse effects mentions in tweets (Task 1) and Generaliz- able identification of personal health expe- rience mentions (Task 4). For our submis- sions, we exploited ensembles based on a pre- trained language representation with a neu- ral transformer architecture (BERT) (Tasks 1 and 4) and a CNN-BiLSTM(-CRF) network within a multi-task learning scenario (Task 1). These systems are placed on top of a carefully crafted pipeline of domain-specific pre- processing steps.
Published in Proceedings of the Fourth Social Media Mining for Health Applications (#smm4h) Workshop & Shared Task, Association for Computational Linguistics, pp. 58–61.
Approaching SMM4H with merged models and multi-task learning
@INPROCEEDINGS{rinaldi2019g,
title = {Approaching {SMM4H} with merged models and multi-task learning},
publisher = {Association for Computational Linguistics},
booktitle = {Proceedings of the Fourth Social Media Mining for Health Applications (\#smm4h) Workshop & Shared Task},
author = {Ellendorff, T. and Furrer, L. and Colic, N. and Aepli, N. and Rinaldi, F.},
pages = {58--61},
year = {2019},
doi = {10.18653/v1/W19-3208},
url = {https://www.aclweb.org/anthology/W19-3208}
}
DownloadFurrer, L., Cornelius, J., Rinaldi, F. (2019). UZH@CRAFT-ST: a sequence-labeling approach to concept recognition. In Proceedings of the 5th Workshop on Bionlp Open Shared Tasks, Association for Computational Linguistics, pp. 185–195.
UZH@CRAFT-ST: a sequence-labeling approach to concept recognition
Authors: Furrer, L. and Cornelius, J. and Rinaldi, F.
Year: 2019
Abstract: As our submission to the CRAFT shared task 2019, we present two neural approaches to concept recognition. We propose two different systems for joint named entity recognition (NER) and normalization (NEN), both of which model the task as a sequence labeling problem. Our first system is a BiLSTM network with two separate outputs for NER and NEN trained from scratch, whereas the second system is an instance of BioBERT fine-tuned on the concept-recognition task. We exploit two strategies for extending concept coverage, ontology pretraining and backoff with a dictionary lookup. Our results show that the backoff strategy effectively tackles the problem of unseen concepts, addressing a major limitation of the chosen design. In the cross-system comparison, BioBERT proves to be a strong basis for creating a concept-recognition system, although some entity types are predicted more accurately by the BiLSTM-based system.
Published in Proceedings of the 5th Workshop on Bionlp Open Shared Tasks, Association for Computational Linguistics, pp. 185–195.
UZH@CRAFT-ST: a sequence-labeling approach to concept recognition
@INPROCEEDINGS{rinaldi2019h,
title = {{UZH@CRAFT}-{ST}: a sequence-labeling approach to concept recognition},
publisher = {Association for Computational Linguistics},
booktitle = {Proceedings of the 5th Workshop on Bionlp Open Shared Tasks},
author = {Furrer, L. and Cornelius, J. and Rinaldi, F.},
pages = {185--195},
year = {2019},
doi = {10.18653/v1/D19-5726},
url = {}
}
DownloadFurrer, L., Jancso, A., Colic, N., Rinaldi, F. (2019). Oger++: hybrid multi-type entity recognition. Journal of Cheminformatics 11(1), 7.
Oger++: hybrid multi-type entity recognition
Authors: Furrer, L. and Jancso, A. and Colic, N. and Rinaldi, F.
Year: 2019
Abstract: Background: We present a text-mining tool for recognizing biomedical entities in scientific literature. OGER++ is a hybrid system for named entity recognition and concept recognition (linking), which combines a dictionary-based annotator with a corpus-based disambiguation component. The annotator uses an efficient look-up strategy combined with a normalization method for matching spelling variants. The disambiguation classifier is implemented as a feed-forward neural network which acts as a postfilter to the previous step. Results: We evaluated the system in terms of processing speed and annotation quality. In the speed benchmarks, the OGER++ web service processes 9.7 abstracts or 0.9 full-text documents per second. On the CRAFT corpus, we achieved 71.4% and 56.7% F1 for named entity recognition and concept recognition, respectively. Conclusions: Combining knowledge-based and data-driven components allows creating a system with competitive performance in biomedical text mining.
Published in Journal of Cheminformatics 11(1), BioMed Central, 7.
Oger++: hybrid multi-type entity recognition
@ARTICLE{rinaldi2019d,
title = {Oger++: hybrid multi-type entity recognition},
journal = {Journal of Cheminformatics},
publisher = {BioMed Central},
volume = {11},
author = {Furrer, L. and Jancso, A. and Colic, N. and Rinaldi, F.},
number = {1},
pages = {7},
year = {2019},
doi = {10.1186/s13321-018-0326-3},
url = {https://doi.org/10.5167/uzh-162875}
}
DownloadKanjirangat, V., Antonucci, A. (2019). NOVEL2GRAPH: Visual summaries of narrative text enhanced by machine learning. Proceedings of the Text2StoryIR'19 Workshop, Cologne, Germany, 14-April-2019, pp. 29–37.
NOVEL2GRAPH: Visual summaries of narrative text enhanced by machine learning
Authors: Kanjirangat, V. and Antonucci, A.
Year: 2019
Abstract: A machine learning approach to the creation of visual summaries for
narrative text is presented. Standard natural language processing tools
for named entities recognition are used together with a clustering algorithm
to detect the characters of the novel and their aliases. The
most relevant ones and their relations are evaluated on the basis of
a simple statistical analysis. These characters are visually depicted
as nodes of an undirected graph whose edges describe relations with
other characters. Specialized sentiment analysis techniques based on
sentence embedding decide the colours of characters/nodes and their
relations/edges. Additional information about the characters (e.g., gender)
and their relations (e.g., siblings or partnerships) are returned by
binary classifiers and visually depicted in the graph. For those specialized
tasks, small amounts of manually annotated data are sucient
to achieve good accuracy. Compared to analogous tools, the machine
learning approach we present allows for a richer representation of texts
of this kind. A case study to demonstrate this approach for a series of
books is also reported.
Published in A. Jorge, R. Campos, A. Jatowt, S. Bhatia (Ed), Proceedings of the Text2StoryIR'19 Workshop, Cologne, Germany, 14-April-2019, ceur, pp. 29–37.
NOVEL2GRAPH: Visual summaries of narrative text enhanced by machine learning
@ARTICLE{vani2019a,
title = {{NOVEL2GRAPH}: {V}isual summaries of narrative text enhanced by machine learning},
journal = {Proceedings of the {Text2StoryIR'19} Workshop, Cologne, Germany, 14-April-2019},
editor = {A. Jorge, R. Campos, A. Jatowt, S. Bhatia},
publisher = {ceur},
author = {Kanjirangat, V. and Antonucci, A.},
pages = {29--37},
year = {2019},
doi = {},
url = {http://ceur-ws.org/Vol-2342/paper4.pdf}
}
DownloadKanjirangat, V., Oita, M., Oezdemir-Zaech, F. (2019). Semantically corroborating neural attention for biomedical question answering. In BioASQ: Large-Scale Biomedical Semantic Indexing and Question Answering: Workshop of ECML/PKDD 2019, Springer, Lecture Notes in Computer Science.
Semantically corroborating neural attention for biomedical question answering
Authors: Kanjirangat, V. and Oita, M. and Oezdemir-Zaech, F.
Year: 2019
Abstract: Biomedical question answering is a great challenge in NLP
due to complex scientific vocabulary and lack of massive annotated corpora, but, at the same time, is full of potential in optimizing in critical ways the biomedical practices. This paper describes the work carried out as a part of the BioASQ challenge (Task-7B Phase-B), and targets an integral step in the question answering process: extractive answer selection.This deals with the identification of the exact answer (words, phrases or sentences) from given article snippets that are related to the question at hand. We address this problem in the context of factoid and summarization question types, using a variety of deep learning and semantic methods, including various architectures (e.g, Dynamic Memory Networks and Bidirectional Attention Flow), transfer learning, biomedical
named entity recognition and corroboration of semantic evidence. On the top of candidate answer selection module, answer prediction to yes/no question types is also addressed by incorporating a sentiment analysis approach. The evaluation with respect to Rouge, MRR and F1 scores, in
relation to the type of question answering task being considered, exhibits the potential of this hybrid method in extracting the correct answer to a question. In addition, the proposed corroborating semantics module can be added on top of the typical QA pipeline to gain a measured 5% improvement in identifying the exact answer with respect to the gold standard.
Accepted in BioASQ: Large-Scale Biomedical Semantic Indexing and Question Answering: Workshop of ECML/PKDD 2019, Springer, Lecture Notes in Computer Science.
Semantically corroborating neural attention for biomedical question answering
@INPROCEEDINGS{supsi2019d,
title = {Semantically corroborating neural attention for biomedical question answering},
publisher = {Springer, Lecture Notes in Computer Science},
booktitle = {{BioASQ}: Large-Scale Biomedical Semantic Indexing and Question Answering: Workshop of {ECML/PKDD} 2019},
author = {Kanjirangat, V. and Oita, M. and Oezdemir-Zaech, F.},
year = {2019},
doi = {},
url = {http://bioasq.org/}
}
DownloadKim, J.D., Cohen, K.B., Collier, N., Lu, Z., Rinaldi, F. (2019). Introduction to BLAH5 special issue: recent progress on interoperability of biomedical text mining. Genomics Inform 17(2), e12.
Introduction to BLAH5 special issue: recent progress on interoperability of biomedical text mining
Authors: Kim, J.D. and Cohen, K.B. and Collier, N. and Lu, Z. and Rinaldi, F.
Year: 2019
Abstract: BLAH is organized annually by the Database Center for Life Science (DBCLS), Research Organization of Information and Systems (ROIS). The goal of the BLAH series is to enhance the interoperability of resources for biomedical text annotation and mining, which we believe is a key for the next breakthrough of biomedical text mining. This special issue delivers seven application notes and two mini reviews, under the theme, “biomedical text mining.” They are outcomes from the 5th Biomedical Linked Annotation Hackathon (BLAH5), which was held from 12th through 15th February 2019 in Kashiwa, Japan.
Published in Genomics Inform 17(2), e12.
Introduction to BLAH5 special issue: recent progress on interoperability of biomedical text mining
@ARTICLE{rinaldi2019b,
title = {Introduction to {BLAH5} special issue: recent progress on interoperability of biomedical text mining},
journal = {Genomics Inform},
volume = {17},
author = {Kim, J.D. and Cohen, K.B. and Collier, N. and Lu, Z. and Rinaldi, F.},
number = {2},
pages = {e12},
year = {2019},
doi = {10.5808/GI.2019.17.2.e12},
url = {}
}
DownloadRenooij, S., van der Gaag, L.C., Leray, Ph. (2019). On intercausal interactions in probabilistic relational models. In De Bock, J., de Campos, C.P., de Cooman, G., Quaeghebeur, E., Wheeler, G. (Eds), Proceedings of the Eleventh International Symposium on Imprecise Probabilities: Theories and Applications (ISIPTA '19), Proceedings of Machine Learning Research 103, pp. 327–329.
On intercausal interactions in probabilistic relational models
Authors: Renooij, S. and van der Gaag, L.C. and Leray, Ph.
Year: 2019
Abstract: Probabilistic relational models (PRMs) extend Bayesian networks beyond propositional expressiveness by allowing the representation of multiple interacting classes. For a specific instance of sets of concrete objects per class, a ground Bayesian network is composed by replicating parts of the PRM. The interactions between the objects that are thereby induced, are not always obvious from the PRM. We demonstrate in this paper that the replicative structure of the ground network in fact constrains the space of possible probability distributions and thereby the possibly patterns of intercausal interaction.
Keywords:
PRM instances, qualitative constraints on probability distributions, intercausal interaction.
Published in De Bock, J., de Campos, C.P., de Cooman, G., Quaeghebeur, E., Wheeler, G. (Eds), Proceedings of the Eleventh International Symposium on Imprecise Probabilities: Theories and Applications (ISIPTA '19), Proceedings of Machine Learning Research 103, pp. 327–329.
On intercausal interactions in probabilistic relational models
@INPROCEEDINGS{Linda2019b,
title = {On intercausal interactions in probabilistic relational models},
editor = {De Bock, J. and de Campos, C.P. and de Cooman, G. and Quaeghebeur, E. and Wheeler, G.},
series = {Proceedings of Machine Learning Research},
volume = {103},
booktitle = {Proceedings of the Eleventh International Symposium on Imprecise Probabilities: Theories and Applications ({ISIPTA} '19)},
author = {Renooij, S. and van der Gaag, L.C. and Leray, Ph.},
pages = {327--329},
year = {2019},
doi = {},
url = {http://www.sipta.org/isipta19/contributions/renooij19.pdf}
}
DownloadMattei, L., Soares, D.L., Antonucci, A., Mauà, D.D., Facchini, A. (2019). Exploring the space of probabilistic sentential decision diagrams. In Proceedings of the 3rd Tractable Probabilistic Modeling Workshop, 36th International Conference on Machine Learning.
Exploring the space of probabilistic sentential decision diagrams
Authors: Mattei, L. and Soares, D.L. and Antonucci, A. and Mauà, D.D. and Facchini, A.
Year: 2019
Abstract: Probabilistic sentential decision diagrams (PS- DDs) are annotated circuits providing a possibly compact specification of joint probability mass functions consistent with a formula over a set of propositional variables. PSDD inference is tractable in the sense that marginal queries can be achieved in linear time with respect to the circuit size by traversal algorithms. Unlike other probabilistic graphical models such as Bayesian networks, the problem of learning the structure for PSDDs received relatively little attention. We discuss some preliminary ideas related to the development of pure likelihood-score-based search methods for the learning of PSDD structures fit- ting a formula and a data set of consistent obser- vations. A sampling algorithm for these models is also provided.
Published in Proceedings of the 3rd Tractable Probabilistic Modeling Workshop, 36th International Conference on Machine Learning.
Note: Accepted for pubblication.
Exploring the space of probabilistic sentential decision diagrams
@INPROCEEDINGS{supsi2019a,
title = {Exploring the space of probabilistic sentential decision diagrams},
booktitle = {Proceedings of the 3rd Tractable Probabilistic Modeling Workshop, 36th International Conference on Machine Learning},
author = {Mattei, L. and Soares, D.L. and Antonucci, A. and Mau\`a, D.D. and Facchini, A.},
year = {2019},
doi = {},
url = {https://drive.google.com/file/d/1PkckPpeLUOP_Oeik_q4MprD6ZJeekqHZ/view}
}
DownloadMauri, A., Lettori, J., Fusi, G., Fausti, D., Mor, M., Braghin, F., Legnani, G., Roveda, L. (2019). Mechanical and control design of an industrial exoskeleton for advanced human empowering in heavy parts manipulation tasks. MDPI Robotics.
Mechanical and control design of an industrial exoskeleton for advanced human empowering in heavy parts manipulation tasks
Authors: Mauri, A. and Lettori, J. and Fusi, G. and Fausti, D. and Mor, M. and Braghin, F. and Legnani, G. and Roveda, L.
Year: 2019
Abstract: Exoskeleton robots are a rising technology in industrial contexts to assist humans in onerous applications. Mechanical and control design solutions are intensively investigated to achieve a high performance human-robot collaboration (e.g., transparency, ergonomics, safety, etc.). However, the most of the investigated solutions involve high-cost hardware, complex design solutions and standard actuation. Moreover, state-of-the-art empowering controllers do not allow for online assistance regulation and do not embed advanced safety rules. In the presented work, an industrial exoskeleton with high payload ratio for lifting and transportation of heavy parts is proposed. A low-cost mechanical design solution is described, exploiting compliant actuation at the shoulder joint to increase safety in human-robot cooperation. A hierarchic model-based controller with embedded safety rules is then proposed (including the modeling of the compliant actuator) to actively assist the human while executing the task. An inner optimal controller is proposed for trajectory tracking, while an outer safety-based fuzzy logic controller is proposed to online deform the task trajectory on the basis of the human’s intention of motion. A gain scheduler is also designed to calculate the inner optimal control gains on the basis of the performed trajectory. Simulations have been performed in order to validate the performance of the proposed device, showing promising results. The prototype is under realization.
Published in MDPI Robotics.
Mechanical and control design of an industrial exoskeleton for advanced human empowering in heavy parts manipulation tasks
@ARTICLE{Roveda2019b,
title = {Mechanical and control design of an industrial exoskeleton for advanced human empowering in heavy parts manipulation tasks},
journal = {{MDPI} Robotics},
author = {Mauri, A. and Lettori, J. and Fusi, G. and Fausti, D. and Mor, M. and Braghin, F. and Legnani, G. and Roveda, L.},
year = {2019},
doi = {10.3390/robotics8030065},
url = {}
}
DownloadMejari, M., Piga, D., Toth, R., Bemporad, A. (2019). Kernelized identification of linear parameter-varying models with linear fractional representation. In 2019 European Control Conference (ecc), Naples, Italy.
Kernelized identification of linear parameter-varying models with linear fractional representation
Authors: Mejari, M. and Piga, D. and Toth, R. and Bemporad, A.
Year: 2019
Abstract: The article presents a method for the identification of Linear Parameter-Varying (LPV) models in a Linear Fractional Representation (LFR), which corresponds to a Linear Time-Invariant (LTI) model connected to a scheduling variable dependency via a feedback path. A two-stage identification approach is proposed. In the first stage, Kernelized Canonical Correlation Analysis (KCCA) is formulated to estimate the state sequence of the underlying LPV model. In the second stage, a non-linear least squares cost function is minimized by employing a coordinate descent algorithm to estimate latent variables characterizing the LFR and the unknown model
matrices of the LTI block by using the state estimates obtained at the first stage. Here, it is assumed that the structure of the scheduling variable dependent block in the feedback path is fixed. For a special case of affine dependence of the model on the feedback block, it is shown that the optimization problem in the second stage reduces to ordinary least-squares followed by a singular value decomposition.
Published in 2019 European Control Conference (ecc), Naples, Italy.
Kernelized identification of linear parameter-varying models with linear fractional representation
@INPROCEEDINGS{piga2019e,
title = {Kernelized identification of linear parameter-varying models with linear fractional representation},
address = {Naples, Italy},
booktitle = {2019 European Control Conference ({e}cc)},
author = {Mejari, M. and Piga, D. and Toth, R. and Bemporad, A.},
year = {2019},
doi = {10.23919/ECC.2019.8796150},
url = {}
}
DownloadMiranda, E., Zaffalon, M. (2019). Compatibility, coherence and the RIP. In Destercke, S.,Denoeux, T., Gil, M. A., Grzegorzewski, P., Hryniewicz, O. (Ed), Uncertainty Modelling in Data Science, Advances in Intelligent Systems and Computing 832, Springer, pp. 166–174.
Compatibility, coherence and the RIP
Authors: Miranda, E. and Zaffalon, M.
Year: 2019
Abstract: We generalise the classical result on the compatibility of marginal, possible non-disjoint, assessments in terms of the running intersection property to the imprecise case, where our beliefs are modelled in terms of sets of desirable gambles. We consider the case where we have unconditional and conditional assessments, and show that the problem can be simplified via a tree decomposition.
Published in Destercke, S.,Denoeux, T., Gil, M. A., Grzegorzewski, P., Hryniewicz, O. (Ed), Uncertainty Modelling in Data Science, Advances in Intelligent Systems and Computing 832, Springer, pp. 166–174.
Note: SMPS 2018: Proceedings of the 9th international conference on Soft Methods in Probability and Statistics
Compatibility, coherence and the RIP
@INCOLLECTION{zaffalon2018a,
title = {Compatibility, coherence and the {RIP}},
editor = {Destercke, S.,Denoeux, T., Gil, M. A., Grzegorzewski, P., Hryniewicz, O.},
publisher = {Springer},
series = {Advances in Intelligent Systems and Computing},
volume = {832},
booktitle = {Uncertainty Modelling in Data Science},
author = {Miranda, E. and Zaffalon, M.},
pages = {166--174},
year = {2019},
doi = {10.1007/978-3-319-97547-4_22},
url = {}
}
DownloadMusumeci, F., Rottondi, C.E.M., Corani, G., Shahkarami, S., Cugini, F., Tornatore, M. (2019). A tutorial on machine learning for failure management in optical networks. Journal of Lightwave Technology 37(16).
A tutorial on machine learning for failure management in optical networks
Authors: Musumeci, F. and Rottondi, C.E.M. and Corani, G. and Shahkarami, S. and Cugini, F. and Tornatore, M.
Year: 2019
Abstract: Failure management plays a role of capital importance in optical networks to avoid service disruptions and to satisfy customers' service level agreements. Machine Learning (ML) promises to revolutionize the (mostly manual and human-driven) approaches in which failure management in optical networks has been traditionally managed, by introducing automated methods for failure prediction, detection, localization and identification. This tutorial provides a gentle introduction to some ML techniques that have been recently applied in the field of optical-network failure management. It then introduces a taxonomy to classify failure-management tasks and discusses possible applications of ML for these failure management tasks. Finally, for a reader interested in more implementative details, we provide a step-by-step description of how to solve a representative example of a practical failure-management task.
Accepted in Journal of Lightwave Technology 37(16).
A tutorial on machine learning for failure management in optical networks
@ARTICLE{corani2019b,
title = {A tutorial on machine learning for failure management in optical networks},
journal = {Journal of Lightwave Technology},
volume = {37},
author = {Musumeci, F. and Rottondi, C.E.M. and Corani, G. and Shahkarami, S. and Cugini, F. and Tornatore, M.},
number = {16},
year = {2019},
doi = {10.1109/JLT.2019.2922586},
url = {}
}
DownloadOita, M. (2019). Reverse engineering creativity into interpretable neural networks. In Future of Information and Communications, Lecture Notes in Networks and Systems 70, pp. 235-247.
Reverse engineering creativity into interpretable neural networks
Authors: Oita, M.
Year: 2019
Abstract: In the field of AI the ultimate goal is to achieve generic intelligence, also called True AI, but which depends on the successful enablement of imagination and creativity in artificial agents. To address this problem, this paper presents a novel deep learning framework for creativity, called INNGenuity. Pursuing an interdisciplinary implementation of creativity conditions, INNGenuity aims at the resolution of the various flaws of current AI learning architectures, which stem from the opacity of their models. Inspired by the neuroanatomy of the brain during creative cognition, the proposed framework's hybrid architecture blends both symbolic and connectionist AI, inline with Minsky's ”society of mind”. At its core, semantic gates are designed to facilitate an input/output flow of semantic structures and enable the usage of aligning mechanisms between neural activation clusters and semantic graphs. Having as goal alignment maximization, such a system would enable interpretability through the creation of labeled patterns of computation, and propose unaligned but relevant computation patterns as novel and useful, therefore creative.
Published in Future of Information and Communications, Lecture Notes in Networks and Systems 70, pp. 235-247.
Reverse engineering creativity into interpretable neural networks
@INPROCEEDINGS{oita2019innGenuity,
title = {Reverse engineering creativity into interpretable neural networks},
series = {Lecture Notes in Networks and Systems},
volume = {70},
booktitle = {Future of Information and Communications},
author = {Oita, M.},
pages = {235-247},
year = {2019},
doi = {10.1007/978-3-030-12385-7_19},
url = {}
}
DownloadOita, M. (2019). Incremental alignment of metaphoric language model for poetry composition. In Computing Conference, Springer, "Advances in Intelligent Systems and Computing", pp. 834–845.
Incremental alignment of metaphoric language model for poetry composition
Authors: Oita, M.
Year: 2019
Abstract: The ability to automatically generate meaningful text with respect to a topic is an important AI mission. In particular, the automatic generation of content which is deemed creative is a great challenge. In this paper, poetry generation is approached through the lenses of a new architectural design for creativity that leverages semantics for the creation of variance and the preservation of the content coherency throughout the generation process. The fully implemented system is made available on github.
Accepted in Computing Conference, Springer, "Advances in Intelligent Systems and Computing", pp. 834–845.
Incremental alignment of metaphoric language model for poetry composition
@INPROCEEDINGS{oita2019poetryComposition,
title = {Incremental alignment of metaphoric language model for poetry composition},
publisher = {Springer, "Advances in Intelligent Systems and Computing"},
booktitle = {Computing Conference},
author = {Oita, M.},
pages = {834--845},
year = {2019},
doi = {10.1007/978-3-030-22871-2_59},
url = {}
}
DownloadPiga, D. (2019). Finite-horizon integration for continuous-time identification: bias analysis and application to variable stiffness actuators. International Journal of Control, pp. 1–14.
Finite-horizon integration for continuous-time identification: bias analysis and application to variable stiffness actuators
Authors: Piga, D.
Year: 2019
Abstract: Direct identification of continuous-time dynamical models from sampled data is now a mature discipline, which is known to have many advantages with respect to indirect approaches based on the identification of discretized models. This paper faces the problem of continuous-time identification of linear time-invariant systems through finite-horizon numerical integration and least-square estimation. The bias in the least-squares estimator due to the noise corrupting the signal observations is quantified, and the benefits of numerical integration in the attenuation of this bias are discussed. An extension of the approach which combines numerical integration, least-squares estimation and particle swarm optimization is proposed for the identification of nonlinear systems and nonlinear-in-the-parameter models, and then applied to the estimation of the torque-displacement characteristic of a commercial variable stiffness actuator driving a one-degree-of-freedom pendulum.
Accepted in International Journal of Control, Taylor & Francis, pp. 1–14.
Finite-horizon integration for continuous-time identification: bias analysis and application to variable stiffness actuators
@ARTICLE{piga2019c,
title = {Finite-horizon integration for continuous-time identification: bias analysis and application to variable stiffness actuators},
journal = {International Journal of Control},
publisher = {Taylor & Francis},
author = {Piga, D.},
pages = {1--14},
year = {2019},
doi = {10.1080/00207179.2018.1557348},
url = {}
}
DownloadPiga, D., Benavoli, A. (2019). Semialgebraic outer approximations for set-valued nonlinear filtering. In 2019 European Control Conference (ECC), Naples, Italy.
Semialgebraic outer approximations for set-valued nonlinear filtering
Authors: Piga, D. and Benavoli, A.
Year: 2019
Abstract: This paper addresses the set-valued filtering problem for discrete time-varying dynamical systems, whose process and measurement equations are polynomial functions of the system state. According to a set-membership framework, the process and measurement noises, as well as the initial state, are assumed to belong to bounded uncertainty regions, which are supposed to be generic semialgebraic sets described by polynomial inequalities. A sequential algorithm, based on sum-of-squares (SOS) representation of positive polynomials is proposed to compute a semialgebraic set described by an a-priori fixed number of polynomial constraints which is guaranteed to contain the true state of the system with certainty.
Published in 2019 European Control Conference (ECC), Naples, Italy.
Semialgebraic outer approximations for set-valued nonlinear filtering
@INPROCEEDINGS{piga2019f,
title = {Semialgebraic outer approximations for set-valued nonlinear filtering},
address = {Naples, Italy},
booktitle = {2019 European Control Conference ({ECC})},
author = {Piga, D. and Benavoli, A.},
year = {2019},
doi = {10.23919/ECC.2019.8795731},
url = {}
}
DownloadPiga, D., Forgione, M., Formentin, S., Bemporad, A. (2019). Performance-oriented model learning for data-driven MPC design. IEEE Control Systems Letters 3(3), pp. 577 - 582.
Performance-oriented model learning for data-driven MPC design
Authors: Piga, D. and Forgione, M. and Formentin, S. and Bemporad, A.
Year: 2019
Abstract: Model Predictive Control (MPC) is an enabling technology in applications requiring controlling physical processes in an optimized way under constraints on inputs and outputs. However, in MPC closed-loop performance is pushed to the limits only if the plant under control is accurately modeled; otherwise, robust architectures need to be employed, at the price of reduced performance due to worst-case conservative assumptions. In this paper, instead of adapting the controller to handle uncertainty, we adapt the learning procedure so that the prediction model is selected to provide the best closed-loop performance. More specifically, we apply for the first time the above “identification for control” rationale to hierarchical MPC using data-driven methods and Bayesian optimization.
Published in IEEE Control Systems Letters 3(3), pp. 577 - 582.
Performance-oriented model learning for data-driven MPC design
@ARTICLE{piga2019a,
title = {Performance-oriented model learning for data-driven {MPC} design},
journal = {{IEEE} Control Systems Letters},
volume = {3},
author = {Piga, D. and Forgione, M. and Formentin, S. and Bemporad, A.},
number = {3},
pages = {577 - 582},
year = {2019},
doi = {10.1109/LCSYS.2019.2913347},
url = {}
}
DownloadRenooij, S., van der Gaag, L.C. (2019). The hidden elegance of causal interaction models. In Ben Amor, N., Quost, B., Theobald, M (Eds), 13th International Conference on Scalable Uncertainty Management (SUM '19), Lecture Notes in Artificial Intelligence 11940, Springer, pp. 38–51.
The hidden elegance of causal interaction models
Authors: Renooij, S. and van der Gaag, L.C.
Year: 2019
Abstract: Causal interaction models such as the noisy-OR model, are used in Bayesian networks to simplify probability acquisition for variables with large numbers of modelled causes. These models essentially prescribe how to complete an exponentially large probability table from a linear number of parameters. Yet, typically the full probability tables are required for inference with Bayesian networks in which such interaction models are used, although inference algorithms tailored to specific types of network exist that can directly exploit the decomposition properties of the interaction models. In this paper we revisit these decomposition properties in view of general inference algorithms and demonstrate that they allow an alternative representation of causal interaction models that is quite concise, even with large numbers of causes involved. In addition to forestalling the need of tailored algorithms, our alternative representation brings engineering benefits beyond those widely recognised.
Keywords: Bayesian networks, causal interaction models, maintenance robustness
Published in Ben Amor, N., Quost, B., Theobald, M (Eds), 13th International Conference on Scalable Uncertainty Management (SUM '19), Lecture Notes in Artificial Intelligence 11940, Springer, pp. 38–51.
The hidden elegance of causal interaction models
@INPROCEEDINGS{linda2019a,
title = {The hidden elegance of causal interaction models},
editor = {Ben Amor, N. and Quost, B. and Theobald, M},
publisher = {Springer},
series = {Lecture Notes in Artificial Intelligence},
volume = {11940},
booktitle = {13th International Conference on Scalable Uncertainty Management ({SUM} '19)},
author = {Renooij, S. and van der Gaag, L.C.},
pages = {38--51},
year = {2019},
doi = {10.1007/978-3-030-35514-2_4},
url = {}
}
DownloadRodriguez-Esteban, R., Vishnyakova, D., Rinaldi, F. (2019). Revisiting the decay of scientific email addresses. bioRxiv.
Revisiting the decay of scientific email addresses
Authors: Rodriguez-Esteban, R. and Vishnyakova, D. and Rinaldi, F.
Year: 2019
Abstract: Email is the primary means of communication for scientists. However, scientific authors change email address over time. Using a new method, we have calculated that approximately 18% of all authorstextquoteright contact email addresses in MEDLINE are invalid. While an unfortunate number, it is, however, lower than previously estimated. To mitigate this problem, institutions should provide email forwarding and scientific authors should use more stable email addresses. In fact, a steadily growing share already use free private email addresses: 32% of all new addresses in MEDLINE in 2018 were of this kind.
Published in bioRxiv, Cold Spring Harbor Laboratory.
Revisiting the decay of scientific email addresses
@ARTICLE{rinaldi2019j,
title = {Revisiting the decay of scientific email addresses},
journal = {{bioRxiv}},
publisher = {Cold Spring Harbor Laboratory},
author = {Rodriguez-Esteban, R. and Vishnyakova, D. and Rinaldi, F.},
year = {2019},
doi = {10.1101/633255},
url = {https://www.biorxiv.org/content/early/2019/05/12/633255}
}
DownloadRoveda, L., Haghshenas, S., Caimmi, M., Pedrocchi, N., Molinari Tosatti, L. (2019). Assisting operators in heavy industrial tasks: on the design of an optimized cooperative impedance fuzzy-controller with embedded safety rules. Frontiers in Robotics and AI 6, 75.
Assisting operators in heavy industrial tasks: on the design of an optimized cooperative impedance fuzzy-controller with embedded safety rules
Authors: Roveda, L. and Haghshenas, S. and Caimmi, M. and Pedrocchi, N. and Molinari Tosatti, L.
Year: 2019
Abstract: Human-robot cooperation is increasingly demanded in industrial applications. Many tasks require the robot to enhance the capabilities of humans. In this scenario, safety also plays an important role in avoiding any accident involving humans, robots, and the environment. With this aim, the paper proposes a cooperative fuzzy-impedance control with embedded safety rules to assist human operators in heavy industrial applications while manipulating unknown weight parts. The proposed methodology is composed by four main components: (i) an inner Cartesian impedance controller (to achieve the compliant robot behavior), (ii) an outer fuzzy controller (to provide the assistance to the human operator), (iii) embedded safety rules (to limit force/velocity during the human-robot interaction enhancing safety), and (iv) a neural network approach (to optimize the control parameters for the human-robot collaboration on the basis of the target indexes of assistance performance defined for this purpose). The main achieved result refers to the capability of the controller to deal with uncertain payloads while assisting and empowering the human operator, both embedding in the controller safety features at force and velocity levels and minimizing the proposed performance indexes. The effectiveness of the proposed approach is verified with a KUKA iiwa 14 R820 manipulator in an experimental procedure where human subjects evaluate the robot performance in a collaborative lifting task of a 10 kg part.
Published in Frontiers in Robotics and AI 6, 75.
Assisting operators in heavy industrial tasks: on the design of an optimized cooperative impedance fuzzy-controller with embedded safety rules
@ARTICLE{Roveda2019a,
title = {Assisting operators in heavy industrial tasks: on the design of an optimized cooperative impedance fuzzy-controller with embedded safety rules},
journal = {Frontiers in Robotics and {AI}},
volume = {6},
author = {Roveda, L. and Haghshenas, S. and Caimmi, M. and Pedrocchi, N. and Molinari Tosatti, L.},
pages = {75},
year = {2019},
doi = {10.3389/frobt.2019.00075},
url = {}
}
DownloadSalani, M., Corbellini, G., Corani, G. (2019). Hybrid heuristic for the optimal design of photovoltaic installations considering mismatch loss effects. Computers & Operations Research 108, pp. 112–120.
Hybrid heuristic for the optimal design of photovoltaic installations considering mismatch loss effects
Authors: Salani, M. and Corbellini, G. and Corani, G.
Year: 2019
Abstract: We consider the Photovoltaic Installation Design Problem (PIDP). In this problem, photovoltaic modules must be connected in “strings” and wired to a set of electronic components. The aim is to minimize installation costs and maximize power production, which is affected by “mismatch losses” caused by non-uniform irradiation (shading) and is also directly related to design decisions. We relate the problem to the known class of location routing problems and we design a route-first/cluster-second heuristic. We propose an efficient machine learning approach to evaluate Photovoltaic (PV) string performances accounting for mismatch losses. We prove that our approach is effective in real-world instances provided by our industrial partner.
Published in Computers & Operations Research 108, pp. 112–120.
Hybrid heuristic for the optimal design of photovoltaic installations considering mismatch loss effects
@ARTICLE{corani2019a,
title = {Hybrid heuristic for the optimal design of photovoltaic installations considering mismatch loss effects},
journal = {Computers & Operations Research},
volume = {108},
author = {Salani, M. and Corbellini, G. and Corani, G.},
pages = {112--120},
year = {2019},
doi = {10.1016/j.cor.2019.04.009},
url = {}
}
DownloadSechidis, K., Azzimonti, L., Pocock, A., Corani, G., Weatherall, J., Brown, G. (2019). Efficient feature selection using shrinkage estimators. Machine Learning 108(8), pp. 1261–1286.
Efficient feature selection using shrinkage estimators
Authors: Sechidis, K. and Azzimonti, L. and Pocock, A. and Corani, G. and Weatherall, J. and Brown, G.
Year: 2019
Abstract: Information theoretic feature selection methods quantify the importance of each feature by estimating mutual information terms to capture: the relevancy, the redundancy and the complementarity. These terms are commonly estimated by maximum likelihood, while an under-explored area of research is how to use shrinkage methods instead. Our work suggests a novel shrinkage method for data-efficient estimation of information theoretic terms. The small sample behaviour makes it particularly suitable for estimation of discrete distributions with large number of categories (bins). Using our novel estimators we derive a framework for generating feature selection criteria that capture any high-order feature interaction for redundancy and complementarity. We perform a thorough empirical study across datasets from diverse sources and using various evaluation measures. Our first finding is that our shrinkage based methods achieve better results, while they keep the same computational cost as the simple maximum likelihood based methods. Furthermore, under our framework we derive efficient novel high-order criteria that outperform state-of-the-art methods in various tasks.
Published in Machine Learning 108(8), pp. 1261–1286.
Efficient feature selection using shrinkage estimators
@ARTICLE{azzimonti2019b,
title = {Efficient feature selection using shrinkage estimators},
journal = {Machine Learning},
volume = {108},
author = {Sechidis, K. and Azzimonti, L. and Pocock, A. and Corani, G. and Weatherall, J. and Brown, G.},
number = {8},
pages = {1261--1286},
year = {2019},
doi = {10.1007/s10994-019-05795-1},
url = {}
}
DownloadSheikhalishahi, S., Miotto, R., Dudley, J.T., Lavelli, A., Rinaldi, F., Osmani, V. (2019). Natural language processing of clinical notes on chronic diseases: systematic review. JMIR Med Inform 7(2), e12239.
Natural language processing of clinical notes on chronic diseases: systematic review
Authors: Sheikhalishahi, S. and Miotto, R. and Dudley, J.T. and Lavelli, A. and Rinaldi, F. and Osmani, V.
Year: 2019
Abstract: Background: Novel approaches that complement and go beyond evidence-based medicine are required in the domain of chronic diseases, given the growing incidence of such conditions on the worldwide population. A promising avenue is the secondary use of electronic health records (EHRs), where patient data are analyzed to conduct clinical and translational research. Methods based on machine learning to process EHRs are resulting in improved understanding of patient clinical trajectories and chronic disease risk prediction, creating a unique opportunity to derive previously unknown clinical insights. However, a wealth of clinical histories remains locked behind clinical narratives in free-form text. Consequently, unlocking the full potential of EHR data is contingent on the development of natural language processing (NLP) methods to automatically transform clinical text into structured clinical data that can guide clinical decisions and potentially delay or prevent disease onset. Objective: The goal of the research was to provide a comprehensive overview of the development and uptake of NLP methods applied to free-text clinical notes related to chronic diseases, including the investigation of challenges faced by NLP methodologies in understanding clinical narratives. Methods: Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed and searches were conducted in 5 databases using “clinical notes,” “natural language processing,” and “chronic disease” and their variations as keywords to maximize coverage of the articles. Results: Of the 2652 articles considered, 106 met the inclusion criteria. Review of the included papers resulted in identification of 43 chronic diseases, which were then further classified into 10 disease categories using the International Classification of Diseases, 10th Revision. The majority of studies focused on diseases of the circulatory system (n=38) while endocrine and metabolic diseases were fewest (n=14). This was due to the structure of clinical records related to metabolic diseases, which typically contain much more structured data, compared with medical records for diseases of the circulatory system, which focus more on unstructured data and consequently have seen a stronger focus of NLP. The review has shown that there is a significant increase in the use of machine learning methods compared to rule-based approaches; however, deep learning methods remain emergent (n=3). Consequently, the majority of works focus on classification of disease phenotype with only a handful of papers addressing extraction of comorbidities from the free text or integration of clinical notes with structured data. There is a notable use of relatively simple methods, such as shallow classifiers (or combination with rule-based methods), due to the interpretability of predictions, which still represents a significant issue for more complex methods. Finally, scarcity of publicly available data may also have contributed to insufficient development of more advanced methods, such as extraction of word embeddings from clinical notes. Conclusions: Efforts are still required to improve (1) progression of clinical NLP methods from extraction toward understanding; (2) recognition of relations among entities rather than entities in isolation; (3) temporal extraction to understand past, current, and future clinical events; (4) exploitation of alternative sources of clinical knowledge; and (5) availability of large-scale, de-identified clinical corpora.
Published in JMIR Med Inform 7(2), e12239.
Natural language processing of clinical notes on chronic diseases: systematic review
@ARTICLE{rinaldi2019f,
title = {Natural language processing of clinical notes on chronic diseases: systematic review},
journal = {{JMIR} Med Inform},
volume = {7},
author = {Sheikhalishahi, S. and Miotto, R. and Dudley, J.T. and Lavelli, A. and Rinaldi, F. and Osmani, V.},
number = {2},
pages = {e12239},
year = {2019},
doi = {10.2196/12239},
url = {http://medinform.jmir.org/2019/2/e12239/}
}
DownloadVishnyakova, D., Rodriguez-Esteban, R., Rinaldi, F. (2019). A new approach and gold standard toward author disambiguation in MEDLINE. J Am Med Inform Assoc 26(10), pp. 1037–1045.
A new approach and gold standard toward author disambiguation in MEDLINE
Authors: Vishnyakova, D. and Rodriguez-Esteban, R. and Rinaldi, F.
Year: 2019
Abstract: Author-centric analyses of fast-growing biomedical reference databases are challenging due to author ambiguity. This problem has been mainly addressed through author disambiguation using supervised machine-learning algorithms. Such algorithms, however, require adequately designed gold standards that reflect the reference database properly. In this study we used MEDLINE to build the first unbiased gold standard in a reference database and improve over the existing state of the art in author disambiguation. Following a new corpus design method, publication pairs randomly picked from MEDLINE were evaluated by both crowdsourcing and expert curators. Because the latter showed higher accuracy than crowdsourcing, expert curators were tasked to create a full corpus. The corpus was then used to explore new features that could improve state-of-the-art author disambiguation algorithms that would not have been discoverable with previously existing gold standards. We created a gold standard based on 1900 publication pairs that shows close similarity to MEDLINE in terms of chronological distribution and information completeness. A machine-learning algorithm that includes new features related to the ethnic origin of authors showed significant improvements over the current state of the art and demonstrates the necessity of realistic gold standards to further develop effective author disambiguation algorithms. An unbiased gold standard can give a more accurate picture of the status of author disambiguation research and help in the discovery of new features for machine learning. The principles and methods shown here can be applied to other reference databases beyond MEDLINE.
Published in J Am Med Inform Assoc 26(10), pp. 1037–1045.
A new approach and gold standard toward author disambiguation in MEDLINE
@ARTICLE{rinaldi2019a,
title = {A new approach and gold standard toward author disambiguation in {MEDLINE}},
journal = {J Am Med Inform Assoc},
volume = {26},
author = {Vishnyakova, D. and Rodriguez-Esteban, R. and Rinaldi, F.},
number = {10},
pages = {1037--1045},
year = {2019},
doi = {10.1093/jamia/ocz028},
url = {}
}
Download top2018
Antonucci, A., Facchini, A. (2018). A credal extension of independent choice logic. In Proceedings of the 12th International Conference on Scalable Uncertainty Management (SUM 2018), pp. 35-49.
A credal extension of independent choice logic
Authors: Antonucci, A. and Facchini, A.
Year: 2018
Abstract: We propose an extension of Poole’s independent choice logic based on a relaxation of the underlying independence assumptions. A credal semantics involving multiple joint probability mass functions over the possible worlds is adopted. This represents a conservative approach to probabilistic logic program- ming achieved by considering all the mass functions consistent with the prob- abilistic facts. This allows to model tasks for which independence among some probabilistic choices cannot be assumed, and a specific dependence model cannot be assessed. Preliminary tests on an object ranking application show that, despite the loose underlying assumptions, informative inferences can be extracted.
Published in Proceedings of the 12th International Conference on Scalable Uncertainty Management (SUM 2018), pp. 35-49.
A credal extension of independent choice logic
@INPROCEEDINGS{antonucci2018c,
title = {A credal extension of independent choice logic},
booktitle = {Proceedings of the 12th International Conference on Scalable Uncertainty Management ({SUM} 2018)},
author = {Antonucci, A. and Facchini, A.},
pages = {35-49},
year = {2018},
doi = {10.1007/978-3-030-00461-3_3},
url = {https://arxiv.org/abs/1806.08298}
}
DownloadAntonucci, A., Facchini, A. (2018). Set-valued probabilistic sentential decision diagrams. In Proceedings of the 5th Workshop on Probabilistic Logic Programming.
Set-valued probabilistic sentential decision diagrams
Authors: Antonucci, A. and Facchini, A.
Year: 2018
Abstract: Probabilistic sentential decision diagrams are a class of arithmetic circuits locally annotated by probability mass functions. This al- lows for a compact representation of joint mass functions in the presence of logical constraints. Here we propose a set-valued generalisation of the probabilistic quantification in these models, that allows to replace the sharp specification of the local probabilities with linear constraints over them. This induces a (convex) set of joint probability mass functions, all consistent with the assigned logical constraints. These models should be adopted for a cautious learning of the local parameters if only small amount of data are available. Algorithmic strategies to compute the lower and upper bounds of marginal and conditional queries with respect to these sets of joint mass functions are sketched. The task can be achieved in linear time with respect to the diagram size for marginal queries and, if the diagram is singly connected, the same can be done for conditional queries.
Published in Proceedings of the 5th Workshop on Probabilistic Logic Programming.
Set-valued probabilistic sentential decision diagrams
@INPROCEEDINGS{antonucci2018d,
title = {Set-valued probabilistic sentential decision diagrams},
booktitle = {Proceedings of the 5th Workshop on Probabilistic Logic Programming},
author = {Antonucci, A. and Facchini, A.},
year = {2018},
doi = {},
url = {http://stoics.org.uk/plp/plp2018/}
}
DownloadBemporad, A., Breschi, V., Piga, D., Boyd, S. (2018). Fitting jump models. Automatica 96, pp. 11–21.
Fitting jump models
Authors: Bemporad, A. and Breschi, V. and Piga, D. and Boyd, S.
Year: 2018
Abstract: We describe a new framework for fitting jump models to a sequence of data. The key idea is to alternate between minimizing a loss function to fit multiple model parameters, and minimizing a discrete loss function to determine which set of model parameters is active at each data point. The framework is quite general and encompasses popular classes of models, such as hidden Markov models and piecewise affine models. The shape of the chosen loss functions to minimize determines the shape of the resulting jump model.
Published in Automatica 96, pp. 11–21.
Fitting jump models
@ARTICLE{piga2018e,
title = {Fitting jump models},
journal = {Automatica},
volume = {96},
author = {Bemporad, A. and Breschi, V. and Piga, D. and Boyd, S.},
pages = {11--21},
year = {2018},
doi = {10.1016/j.automatica.2018.06.022},
url = {}
}
DownloadBreschi, V., Bemporad, A., Piga, D., Boyd, S. (2018). Prediction error methods in learning jump ARMAX models. In 2018 IEEE Conference on Decision and Control (cdc), pp. 2247–2252.
Prediction error methods in learning jump ARMAX models
Authors: Breschi, V. and Bemporad, A. and Piga, D. and Boyd, S.
Year: 2018
Published in 2018 IEEE Conference on Decision and Control (cdc), pp. 2247–2252.
Prediction error methods in learning jump ARMAX models
@INPROCEEDINGS{piga2018i,
title = {Prediction error methods in learning jump {ARMAX} models},
booktitle = {2018 {IEEE} Conference on Decision and Control ({c}dc)},
author = {Breschi, V. and Bemporad, A. and Piga, D. and Boyd, S.},
pages = {2247--2252},
year = {2018},
doi = {10.1109/CDC.2018.8619819},
url = {}
}
DownloadBreschi, V., Piga, D., Bemporad, A. (2018). Kalman filtering for energy disaggregation. In Proc. of the 1st IFAC Workshop on Integrated Assessment Modelling for Environmental Systems 51(5), pp. 108–113.
Kalman filtering for energy disaggregation
Authors: Breschi, V. and Piga, D. and Bemporad, A.
Year: 2018
Abstract: Providing the users information on the energy consumed in the household at the appliance level is of major importance for increasing their awareness of their consumption behavior. In this paper, we propose a technique based on Kalman filters to estimate the devices' consumption patterns from aggregate readings, i.e., to solve the so called disaggregation problem. The method is suited for on-line disaggregation and the proposed results show that it is robust against modelling errors and unmodelled appliances.
Published in Proc. of the 1st IFAC Workshop on Integrated Assessment Modelling for Environmental Systems IFAC-PapersOnLine 51(5), pp. 108–113.
Kalman filtering for energy disaggregation
@INPROCEEDINGS{piga2018b,
title = {Kalman filtering for energy disaggregation},
journal = {{IFAC}-{PapersOnLine}},
volume = {51},
booktitle = {Proc. {o}f the 1st {IFAC} Workshop on Integrated Assessment Modelling for Environmental Systems},
author = {Breschi, V. and Piga, D. and Bemporad, A.},
number = {5},
pages = {108--113},
year = {2018},
doi = {10.1016/j.ifacol.2018.06.219},
url = {}
}
DownloadBreschi, V., Piga, D., Bemporad, A. (2018). Jump model learning and filtering for energy end-use disaggregation. In Proc. of the 18th IFAC Symposium on System Identification 51(15), pp. 275–280.
Jump model learning and filtering for energy end-use disaggregation
Authors: Breschi, V. and Piga, D. and Bemporad, A.
Year: 2018
Abstract: Energy disaggregation aims at reconstructing the power consumed by each electric appliance available in a household from the aggregate power readings collected by a single-point smart meter. With the ultimate goal of fully automatizing this procedure, we first estimate a set of jump models, each of them describing the consumption behaviour of each electric appliance. By representing the total power consumed at the household level as the sum of the outputs of the estimated jump models, a filtering algorithm, based on dynamic programming, is then employed to reconstruct, in an iterative way, the power consumption at an individual appliance level.
Published in Proc. of the 18th IFAC Symposium on System Identification 51(15), pp. 275–280.
Jump model learning and filtering for energy end-use disaggregation
@INPROCEEDINGS{piga2018f,
title = {Jump model learning and filtering for energy end-use disaggregation},
volume = {51},
booktitle = {Proc. {o}f the 18th {IFAC} Symposium on System Identification},
author = {Breschi, V. and Piga, D. and Bemporad, A.},
number = {15},
pages = {275--280},
year = {2018},
doi = {10.1016/j.ifacol.2018.09.147},
url = {}
}
Downloadde Campos, C.P., Scanagatta, M., Corani, G., Zaffalon, M. (2018). Entropy-based pruning for learning Bayesian networks using BIC. Artificial Intelligence 260, pp. 42-50.
Entropy-based pruning for learning Bayesian networks using BIC
Authors: de Campos, C.P. and Scanagatta, M. and Corani, G. and Zaffalon, M.
Year: 2018
Abstract: For decomposable score-based structure learning of Bayesian networks, existing approaches first compute a collection of candidate parent sets for each variable and then optimize over this collection by choosing one parent set for each variable without creating directed cycles while maximizing the total score. We target the task of constructing the collection of candidate parent sets when the score of choice is the Bayesian Information Criterion (BIC). We provide new non-trivial results that can be used to prune the search space of candidate parent sets of each node. We analyze how these new results relate to previous ideas in the literature both theoretically and empirically. We show in experiments with UCI data sets that gains can be significant. Since the new pruning rules are easy to implement and have low computational costs, they can be promptly integrated into all state-of-the-art methods for structure learning of Bayesian networks.
Published in Artificial Intelligence 260, pp. 42-50.
Entropy-based pruning for learning Bayesian networks using BIC
@ARTICLE{deCampos2018a,
title = {Entropy-based pruning for learning {B}ayesian networks using {BIC}},
journal = {Artificial Intelligence},
volume = {260},
author = {de Campos, C.P. and Scanagatta, M. and Corani, G. and Zaffalon, M.},
pages = {42-50},
year = {2018},
doi = {10.1016/j.artint.2018.04.002},
url = {}
}
DownloadKern, H., Corani, G., Huber, D., Vermes, N., Zaffalon, M. (2018). What interplay of factors influences the place of death in cancer patients? an innovative probabilistic approach sheds light on a well-known question. Journal of Pain and Symptom Management 56(6), e25.
What interplay of factors influences the place of death in cancer patients? an innovative probabilistic approach sheds light on a well-known question
Authors: Kern, H. and Corani, G. and Huber, D. and Vermes, N. and Zaffalon, M.
Year: 2018
Abstract: Most terminally ill cancer patients would prefer to die at home although a majority die in institutional settings. Questions about this discrepancy are not yet fully answered. This study applies machine learning techniques to explore the complicated network of factors and the cause-effect relationships affecting the place of death with the ultimate aim of developing policies.
Published in Journal of Pain and Symptom Management Journal of Pain and Symptom Management 56(6), Elsevier, e25.
What interplay of factors influences the place of death in cancer patients? an innovative probabilistic approach sheds light on a well-known question
@ARTICLE{kern2018a,
title = {What interplay of factors influences the place of death in cancer patients? {a}n innovative probabilistic approach sheds light on a well-known question},
journal = {Journal of Pain and Symptom Management},
publisher = {Elsevier},
volume = {56},
booktitle = {Journal of Pain and Symptom Management},
author = {Kern, H. and Corani, G. and Huber, D. and Vermes, N. and Zaffalon, M.},
number = {6},
pages = {e25},
year = {2018},
doi = {10.1016/j.jpainsymman.2018.10.016},
url = {}
}
DownloadMarchetti, S., Antonucci, A. (2018). Reliable uncertain evidence modeling in Bayesian networks by credal networks. In Proceedings of the 31st International Flairs Conference (FLAIRS-31), AAAI Press, pp. 513–518.
Reliable uncertain evidence modeling in Bayesian networks by credal networks
Authors: Marchetti, S. and Antonucci, A.
Year: 2018
Abstract: A reliable modeling of uncertain evidence in Bayesian networks based on a set-valued quantification is proposed. Both soft and virtual evidences are considered. We show that evidence propagation in this setup can be reduced to standard updating in an augmented credal network, equivalent to a set of consistent Bayesian networks. A characterization of the computational complexity for this task is derived together with an efficient exact procedure for a subclass of instances. In the case of multiple uncertain evidences over the same variable, the proposed procedure can provide a set-valued version of the geometric approach to opinion pooling.
Published in Proceedings of the 31st International Flairs Conference (FLAIRS-31), AAAI Press, pp. 513–518.
Reliable uncertain evidence modeling in Bayesian networks by credal networks
@INPROCEEDINGS{antonucci2018a,
title = {Reliable uncertain evidence modeling in {B}ayesian networks by credal networks},
publisher = {AAAI Press},
booktitle = {Proceedings of the 31st International Flairs Conference ({FLAIRS}-31)},
author = {Marchetti, S. and Antonucci, A.},
pages = {513--518},
year = {2018},
doi = {},
url = { https://aaai.org/ocs/index.php/FLAIRS/FLAIRS18/paper/download/17696/16792}
}
DownloadMarchetti, S., Antonucci, A. (2018). Imaginary kinematics. In Proceedings of the 34th Conference on Uncertainty in Artificial Intelligence, AUAI Press, pp. 104-113.
Imaginary kinematics
Authors: Marchetti, S. and Antonucci, A.
Year: 2018
Abstract: We introduce a novel class of adjustment rules for a collection of beliefs. This is an extension of Lewis’ imaging to absorb probabilistic evidence in generalized settings. Unlike standard tools for belief revision, our proposal may be used when information is inconsistent with an agent’s belief base. We show that the functionals we introduce are based on the imaginary counterpart of probability kinematics for standard belief revision, and prove that, under certain conditions, all standard postulates for belief revision are satisfied.
Published in Proceedings of the 34th Conference on Uncertainty in Artificial Intelligence, AUAI Press, pp. 104-113.
Imaginary kinematics
@INPROCEEDINGS{antonucci2018b,
title = {Imaginary kinematics},
publisher = {AUAI Press},
booktitle = {Proceedings of the 34th Conference on Uncertainty in Artificial Intelligence},
author = {Marchetti, S. and Antonucci, A.},
pages = {104-113},
year = {2018},
doi = {},
url = {http://auai.org/uai2018/proceedings/papers/42.pdf}
}
DownloadMejari, M., Naik, V.V., Piga, D., Bemporad, A. (2018). Regularized moving-horizon PWA regression for LPV system identification. In Proc. of the 18th IFAC Symposium on System Identification 51(15), pp. 1092–1097.
Regularized moving-horizon PWA regression for LPV system identification
Authors: Mejari, M. and Naik, V.V. and Piga, D. and Bemporad, A.
Year: 2018
Abstract: This paper addresses the identification of Linear Parameter-Varying (LPV) models through regularized moving-horizon PieceWise Affine (PWA) regression. Specifically, the scheduling-variable space is partitioned into polyhedral regions, where each region is assigned to a PWA function describing the local affine dependence of the LPV model coefficients on the scheduling variable. The regression approach consists of two stages. In the first stage, the data samples are processed iteratively, and a Mixed-Integer Quadratic Programming (MIQP) problem is solved to cluster the scheduling variable observations and simultaneously fit the model parameters to the training data, within a relatively short moving-horizon window of the past. At the second stage, the polyhedral partition of the scheduling-variable space is computed by separating the estimated clusters through linear multi-category discrimination.
Published in Proc. of the 18th IFAC Symposium on System Identification 51(15), pp. 1092–1097.
Note: 18th IFAC Symposium on System Identification SYSID 2018
Regularized moving-horizon PWA regression for LPV system identification
@INPROCEEDINGS{piga2018d,
title = {Regularized moving-horizon {PWA} regression for {LPV} system identification},
volume = {51},
booktitle = {Proc. {o}f the 18th {IFAC} Symposium on System Identification},
author = {Mejari, M. and Naik, V.V. and Piga, D. and Bemporad, A.},
number = {15},
pages = {1092--1097},
year = {2018},
doi = {10.1016/j.ifacol.2018.09.048},
url = {}
}
DownloadMejari, M., Naik, V.V., Piga, D., Bemporad, A. (2018). Energy disaggregation using piecewise affine regression and binary quadratic programming. In 2018 IEEE Conference on Decision and Control (cdc), pp. 3116–3121.
Energy disaggregation using piecewise affine regression and binary quadratic programming
Authors: Mejari, M. and Naik, V.V. and Piga, D. and Bemporad, A.
Year: 2018
Published in 2018 IEEE Conference on Decision and Control (cdc), pp. 3116–3121.
Energy disaggregation using piecewise affine regression and binary quadratic programming
@INPROCEEDINGS{piga2018h,
title = {Energy disaggregation using piecewise affine regression and binary quadratic programming},
booktitle = {2018 {IEEE} Conference on Decision and Control ({c}dc)},
author = {Mejari, M. and Naik, V.V. and Piga, D. and Bemporad, A.},
pages = {3116--3121},
year = {2018},
doi = {10.1109/CDC.2018.8619175},
url = {}
}
DownloadMejari, M., Piga, D., Bemporad, A. (2018). A bias-correction method for closed-loop identification of linear parameter-varying systems. Automatica 87, pp. 128–141.
A bias-correction method for closed-loop identification of linear parameter-varying systems
Authors: Mejari, M. and Piga, D. and Bemporad, A.
Year: 2018
Abstract: Due to safety constraints and unstable open-loop dynamics, system identification of many real-world processes often requiresgathering data from closed-loop experiments. In this paper, we present a bias-correction scheme for closed-loop identification of Linear Parameter-Varying Input–Output (LPV-IO) models, which aims at correcting the bias caused by the correlation between the input signal exciting the process and output noise. The proposed identification algorithm provides a consistent estimate of the open-loop model parameters when both the output signal and the scheduling variable are corrupted by measurement noise. The effectiveness of the proposed methodology is tested in two simulation case studies.
Published in Automatica 87, pp. 128–141.
A bias-correction method for closed-loop identification of linear parameter-varying systems
@ARTICLE{piga2018c,
title = {A bias-correction method for closed-loop identification of linear parameter-varying systems},
journal = {Automatica},
volume = {87},
author = {Mejari, M. and Piga, D. and Bemporad, A.},
pages = {128--141},
year = {2018},
doi = {10.1016/j.automatica.2017.09.014},
url = {}
}
DownloadPiga, D., Formentin, S., Bemporad, A. (2018). Direct data-driven control of constrained systems. IEEE Transactions on Control Systems Technology 26(4), pp. 1422–1429.
Direct data-driven control of constrained systems
Authors: Piga, D. and Formentin, S. and Bemporad, A.
Year: 2018
Abstract: In model-based control design, one often has to describe the plant by a linear model. Deriving such a model poses issues of parameterization, estimation, and validation of the model before designing the controller. In this paper, a direct data-driven control method is proposed for designing controllers that can handle constraints without deriving a model of the plant and directly from data. A hierarchical control architecture is used, in which an inner linear time-invariant for linear parameter-varying controller is first designed to match a simple and a priori specified closed-loop model. Then, an outer model predictive controller is synthesized to handle input/output constraints and to enhance the performance of the inner loop. The effectiveness of the approach is illustrated by means of a simulation and an experimental example. Practical implementation issues are also discussed.
Published in IEEE Transactions on Control Systems Technology 26(4), pp. 1422–1429.
Direct data-driven control of constrained systems
@ARTICLE{piga2018a,
title = {Direct data-driven control of constrained systems},
journal = {{IEEE} Transactions on Control Systems Technology},
volume = {26},
author = {Piga, D. and Formentin, S. and Bemporad, A.},
number = {4},
pages = {1422--1429},
year = {2018},
doi = {10.1109/TCST.2017.2702118},
url = {}
}
DownloadScanagatta, M., Corani, G., de Campos, C.P., Zaffalon, M. (2018). Approximate structure learning for large Bayesian networks. Machine Learning 107(8-10), pp. 1209–1227.
Approximate structure learning for large Bayesian networks
Authors: Scanagatta, M. and Corani, G. and de Campos, C.P. and Zaffalon, M.
Year: 2018
Abstract: We present approximate structure learning algorithms for Bayesian networks. We discuss on the two main phases of the task: the preparation of the cache of the scores and structure optimization, both with bounded and unbounded treewidth. We improve on state-of-the-art methods that rely on an ordering-based search by sampling more effectively the space of the orders. This allows for a remarkable improvement in learning Bayesian networks from thousands of variables.
We also present a thorough study of the accuracy and the running time of inference, comparing bounded-treewidth and unbounded-treewidth models.
Published in Machine Learning 107(8-10), Springer, pp. 1209–1227.
Approximate structure learning for large Bayesian networks
@ARTICLE{scanagatta2018b,
title = {Approximate structure learning for large {B}ayesian networks},
journal = {Machine Learning},
publisher = {Springer},
volume = {107},
author = {Scanagatta, M. and Corani, G. and de Campos, C.P. and Zaffalon, M.},
number = {8-10},
pages = {1209--1227},
year = {2018},
doi = {10.1007/s10994-018-5701-9},
url = {}
}
DownloadScanagatta, M., Corani, G., Zaffalon, M., Yoo, J., Kang, U. (2018). Efficient learning of bounded-treewidth Bayesian networks from complete and incomplete data sets. International Journal of Approximate Reasoning 95, pp. 152–166.
Efficient learning of bounded-treewidth Bayesian networks from complete and incomplete data sets
Authors: Scanagatta, M. and Corani, G. and Zaffalon, M. and Yoo, J. and Kang, U.
Year: 2018
Abstract: Learning a Bayesian networks with bounded treewidth is important for reducing the complexity of the inferences. We present a novel anytime algorithm (k-MAX) method for this task, which scales up to thousands of variables. Through
extensive experiments we show that it consistently yields higher-scoring structures than its competitors on complete data sets. We then consider the problem
of structure learning from incomplete data sets. This can be addressed by structural EM, which however is computationally very demanding. We thus adopt
the novel k-MAX algorithm in the maximization step of structural EM, obtaining an efficient computation of the expected sufficient statistics. We test the
resulting structural EM method on the task of imputing missing data, comparing it against the state-of-the-art approach based on random forests. Our approach achieves the same imputation accuracy of the competitors, but in about
one tenth of the time. Furthermore we show that it has worst-case complexity linear in the input size, and that it is easily parallelizable.
Published in International Journal of Approximate Reasoning 95, pp. 152–166.
Efficient learning of bounded-treewidth Bayesian networks from complete and incomplete data sets
@ARTICLE{scanagatta2018a,
title = {Efficient learning of bounded-treewidth {B}ayesian networks from complete and incomplete data sets},
journal = {International Journal of Approximate Reasoning},
volume = {95},
author = {Scanagatta, M. and Corani, G. and Zaffalon, M. and Yoo, J. and Kang, U.},
pages = {152--166},
year = {2018},
doi = {10.1016/j.ijar.2018.02.004},
url = {}
}
DownloadSelvi, D., Piga, D., Bemporad, A. (2018). Towards direct data-driven model-free design of optimal controllers. In 2018 European Control Conference (ecc), pp. 2836–2841.
Towards direct data-driven model-free design of optimal controllers
Authors: Selvi, D. and Piga, D. and Bemporad, A.
Year: 2018
Abstract: The most critical step in modern direct data-driven control design approaches, such as virtual reference feedback tuning and non-iterative correlation-based tuning, is the choice of an adequate closed-loop reference model. Indeed, the chosen reference model should reflect the desired closed-loop performance but also be reproducible by the underlying unknown process when in closed loop with the synthesized controller. In this paper, we propose a novel approach to compute, directly from data, an “optimal” reference model along with the corresponding controller. The performance index used to define the optimality of the reference model measures the tracking error and the actuator efforts (as it is typical in performance-driven controllers such as linear-quadratic Gaussian control and model predictive control), along with a term penalizing the expected mismatch between the reference model and the actual closed-loop system. The performance index depends on the variables used to parametrize the reference model and the controller, which are optimized through a suitable combination of particle swarm optimization and virtual reference feedback tuning.
Published in 2018 European Control Conference (ecc), pp. 2836–2841.
Towards direct data-driven model-free design of optimal controllers
@INPROCEEDINGS{piga2018g,
title = {Towards direct data-driven model-free design of optimal controllers},
booktitle = {2018 European Control Conference ({e}cc)},
author = {Selvi, D. and Piga, D. and Bemporad, A.},
pages = {2836--2841},
year = {2018},
doi = {10.23919/ECC.2018.8550184},
url = {}
}
Download top2017
Arnone, E., Azzimonti, L., Nobile, F., Sangalli, L.M. (2017). A time-dependent PDE regularization to model functional data defined over spatio-temporal domains. In Aneiros G., Bongiorno E.G., Cao R., Vieu P. (Ed), Functional Statistics and Related Fields, Springer International Publishing, pp. 41–44.
A time-dependent PDE regularization to model functional data defined over spatio-temporal domains
Authors: Arnone, E. and Azzimonti, L. and Nobile, F. and Sangalli, L.M.
Year: 2017
Abstract: We propose a method for the analysis of functional data defined over spatio-temporal domains when prior knowledge on the phenomenon under study is available. The model is based on regression with Partial Differential Equations (PDE) penalization. The PDE formalizes the information on the phenomenon and models the regularity of the field in space and time.
Published in Aneiros G., Bongiorno E.G., Cao R., Vieu P. (Ed), Functional Statistics and Related Fields, Springer International Publishing, pp. 41–44.
A time-dependent PDE regularization to model functional data defined over spatio-temporal domains
@INBOOK{azzimonti2017b,
title = {A time-dependent {PDE} regularization to model functional data defined over spatio-temporal domains},
editor = {Aneiros G., Bongiorno E.G., Cao R., Vieu P. },
publisher = {Springer International Publishing},
booktitle = {Functional Statistics and Related Fields},
author = {Arnone, E. and Azzimonti, L. and Nobile, F. and Sangalli, L.M.},
pages = {41--44},
year = {2017},
doi = {10.1007/978-3-319-55846-2_6},
url = {}
}
DownloadAzzimonti, L., Corani, G., Zaffalon, M. (2017). Hierarchical Multinomial-Dirichlet model for the estimation of conditional probability tables. In Raghavan, V., Aluru, S., Karypis, G., Miele, L., Wu, X. (Ed), 2017 IEEE 17th International Conference on Data Mining (ICDM), pp. 739–744.
Hierarchical Multinomial-Dirichlet model for the estimation of conditional probability tables
Authors: Azzimonti, L. and Corani, G. and Zaffalon, M.
Year: 2017
Abstract: We present a novel approach for estimating conditional probability tables, based on a joint, rather than independent, estimate of the conditional distributions belonging to the same table. We derive exact analytical expressions for the estimators and we analyse their properties both analytically and via simulation. We then apply this method to the estimation of parameters in a Bayesian network. Given the structure of the network, the proposed approach better estimates the joint distribution and significantly improves the classification performance with respect to traditional approaches.
Published in Raghavan, V., Aluru, S., Karypis, G., Miele, L., Wu, X. (Ed), 2017 IEEE 17th International Conference on Data Mining (ICDM), pp. 739–744.
Hierarchical Multinomial-Dirichlet model for the estimation of conditional probability tables
@INPROCEEDINGS{azzimonti2017c,
title = {Hierarchical {M}ultinomial-{D}irichlet model for the estimation of conditional probability tables},
editor = {Raghavan, V., Aluru, S., Karypis, G., Miele, L., Wu, X.},
booktitle = {2017 {IEEE} 17th International Conference on Data Mining ({ICDM})},
author = {Azzimonti, L. and Corani, G. and Zaffalon, M.},
pages = {739--744},
year = {2017},
doi = {10.1109/ICDM.2017.85},
url = {}
}
DownloadBalleri, A., Farina, A., Benavoli, A. (2017). Coordination of optimal guidance law and adaptive radiated waveform for interception and rendezvous problems. IET Radar, Sonar & Navigation 11(7), pp. 1132 - 139.
Coordination of optimal guidance law and adaptive radiated waveform for interception and rendezvous problems
Authors: Balleri, A. and Farina, A. and Benavoli, A.
Year: 2017
Abstract: We present an algorithm that allows an interceptor aircraft equipped with an airborne radar to meet another air target (the intercepted) by developing a guidance law and automatically adapting and optimising the transmitted waveform on a pulse to pulse basis. The algorithm uses a Kalman filter to predict the relative position and speed of the interceptor with respect to the target. The transmitted waveform is automatically selected based on its ambiguity function and accuracy properties along the approaching path. For each pulse, the interceptor predicts its position and velocity with respect to the target, takes a measurement of range and radial velocity and, with the Kalman filter, refines the relative range and range rate estimates. These are fed into a Linear Quadratic Gaussian (LQG) controller that ensures the interceptor reaches the target automatically and successfully with minimum error and with the minimum guidance energy consumption.
Published in IET Radar, Sonar & Navigation 11(7), Institution of Engineering and Technology, pp. 1132 - 139.
Coordination of optimal guidance law and adaptive radiated waveform for interception and rendezvous problems
@ARTICLE{benavoli2017a,
title = {Coordination of optimal guidance law and adaptive radiated waveform for interception and rendezvous problems},
journal = {{IET} Radar, Sonar & Navigation},
publisher = {Institution of Engineering and Technology},
volume = {11},
author = {Balleri, A. and Farina, A. and Benavoli, A.},
number = {7},
pages = {1132 - 139},
year = {2017},
doi = {10.1049/iet-rsn.2016.0547},
url = {}
}
DownloadBenavoli, A., Corani, G., Demsar, J., Zaffalon, M. (2017). Time for a change: a tutorial for comparing multiple classifiers through Bayesian analysis. Journal of Machine Learning Research 18(77), pp. 1–36.
Time for a change: a tutorial for comparing multiple classifiers through Bayesian analysis
Authors: Benavoli, A. and Corani, G. and Demsar, J. and Zaffalon, M.
Year: 2017
Abstract: The machine learning community adopted the use of null hypothesis significance testing (NHST) in order to ensure the statistical validity of results. Many scientific fields however realized the shortcomings of frequentist reasoning and in the most radical cases even banned its use in publications. We should do the same: just as we have embraced the Bayesian paradigm in the development of new machine learning methods, so we should also use it in the analysis of our own results. We argue for abandonment of NHST by exposing its fallacies and, more importantly, offer better - more sound and useful - alternatives for it.
Published in Journal of Machine Learning Research 18(77), pp. 1–36.
Time for a change: a tutorial for comparing multiple classifiers through Bayesian analysis
@ARTICLE{benavoli2016e,
title = {Time for a change: a tutorial for comparing multiple classifiers through {B}ayesian analysis},
journal = {Journal of Machine Learning Research},
volume = {18},
author = {Benavoli, A. and Corani, G. and Demsar, J. and Zaffalon, M.},
number = {77},
pages = {1--36},
year = {2017},
doi = {},
url = {http://jmlr.org/papers/v18/16-305.html}
}
DownloadBenavoli, A., Facchini, A., Vicente-Perez, J., Zaffalon, M. (2017). A polarity theory for sets of desirable gambles. In Proc. ISIPTA '17 Int. Symposium on Imprecise Probability: Theories and Applications, PMLR, pp. 1-12.
A polarity theory for sets of desirable gambles
Authors: Benavoli, A. and Facchini, A. and Vicente-Perez, J. and Zaffalon, M.
Year: 2017
Abstract: Coherent sets of almost desirable gambles and credal sets are known to be equivalent models. That is, there exists a bijection between the two collections of sets preserving the usual operations, e.g. conditioning. Such a correspondence is based on the polarity theory for closed convex cones. Learning from this simple observation, in this paper we introduce a new (lexicographic) polarity theory for general convex cones and then we apply it in order to establish an analogous correspondence between coherent sets of desirable gambles and convex sets of lexicographic probabilities.
Published in Proc. ISIPTA '17 Int. Symposium on Imprecise Probability: Theories and Applications, PMLR, pp. 1-12.
A polarity theory for sets of desirable gambles
@INPROCEEDINGS{Benavoli2017c,
title = {A polarity theory for sets of desirable gambles},
publisher = {PMLR},
booktitle = {Proc. {ISIPTA} '17 Int. Symposium on Imprecise Probability: Theories and Applications},
author = {Benavoli, A. and Facchini, A. and Vicente-Perez, J. and Zaffalon, M.},
pages = {1-12},
year = {2017},
doi = {},
url = {http://proceedings.mlr.press/v62/benavoli17b/benavoli17b.pdf}
}
DownloadBenavoli, A., Facchini, A., Piga, D., Zaffalon, M. (2017). SOS for bounded rationality. In Proceedings of Machine Learning Research 62, PMLR, pp. 25–36.
SOS for bounded rationality
Authors: Benavoli, A. and Facchini, A. and Piga, D. and Zaffalon, M.
Year: 2017
Abstract: In the gambling foundation of probability theory, rationality requires that a subject should always (never) find desirable all nonnegative (negative) gambles, because no matter the result of the exper- iment the subject never (always) decreases her money. Evaluating the nonnegativity of a gamble in infinite spaces is a difficult task. In fact, even if we restrict the gambles to be polynomials in R n , the problem of determining nonnegativity is NP-hard. The aim of this paper is to develop a computable theory of desirable gambles. Instead of requiring the subject to accept all nonnegative gambles, we only require her to accept gambles for which she can efficiently determine the nonnegativity (in particular SOS polynomials). We call this new criterion bounded rationality.
Published in Proceedings of Machine Learning Research 62, PMLR, pp. 25–36.
Note: ISIPTA '17 Int. Symposium on Imprecise Probability: Theories and Applications
SOS for bounded rationality
@INPROCEEDINGS{Benavoli2017b,
title = {{SOS} for bounded rationality},
publisher = {PMLR},
volume = {62},
booktitle = {Proceedings of Machine Learning Research},
author = {Benavoli, A. and Facchini, A. and Piga, D. and Zaffalon, M.},
pages = {25--36},
year = {2017},
doi = {},
url = {http://proceedings.mlr.press/v62/benavoli17a/benavoli17a.pdf}
}
DownloadBenavoli, A., Facchini, A., Zaffalon, M. (2017). Bayes + Hilbert = Quantum Mechanics. In Proceedings of the 14th Interational Conference on Quantum Physics and Logic (qpl 2017), Nijmegen, the Netherlands, 3-7 July.
Bayes + Hilbert = Quantum Mechanics
Authors: Benavoli, A. and Facchini, A. and Zaffalon, M.
Year: 2017
Published in Proceedings of the 14th Interational Conference on Quantum Physics and Logic (qpl 2017), Nijmegen, the Netherlands, 3-7 July.
Bayes + Hilbert = Quantum Mechanics
@INPROCEEDINGS{Benavoli2017m,
title = {Bayes + {H}ilbert = {Q}uantum {M}echanics},
booktitle = {Proceedings of the 14th Interational Conference on Quantum Physics and Logic ({q}pl 2017), Nijmegen, the Netherlands, 3-7 July},
author = {Benavoli, A. and Facchini, A. and Zaffalon, M.},
year = {2017},
doi = {},
url = {http://qpl.science.ru.nl/papers/QPL_2017_paper_4.pdf}
}
DownloadBucher, D., Mangili, F., Bonesana, C., Jonietz, D., Cellina, F., Raubal, M. (2017). Demo abstract: extracting eco-feedback information from automatic activity tracking to promote energy-efficient individual mobility behavior. In 33(1), pp. 1-2.
Demo abstract: extracting eco-feedback information from automatic activity tracking to promote energy-efficient individual mobility behavior
Authors: Bucher, D. and Mangili, F. and Bonesana, C. and Jonietz, D. and Cellina, F. and Raubal, M.
Year: 2017
Abstract: Nowadays, most people own a smartphone which
is well suited to constantly record the movement of its user.
One use of the gathered mobility data is to provide users
with feedback and suggestions for personal behavior change.
Such eco-feedback on mobility patterns may stimulate users
to adopt more energy-efficient mobility choices. In this paper,
we present a methodology to extract mobility patterns from
users’ trajectories, compute alternative transport options, and
aggregate and present them in an intuitive way. The resulting
eco-feedback helps people understand their mobility choices
and explore sustainable alternatives.
Published in Computer Science - Research and Development 33(1), pp. 1-2.
Demo abstract: extracting eco-feedback information from automatic activity tracking to promote energy-efficient individual mobility behavior
@INPROCEEDINGS{mangili2017c,
title = {Demo abstract: extracting eco-feedback information from automatic activity tracking to promote energy-efficient individual mobility behavior},
journal = {Computer Science - Research and Development},
volume = {33},
author = {Bucher, D. and Mangili, F. and Bonesana, C. and Jonietz, D. and Cellina, F. and Raubal, M.},
number = {1},
pages = {1-2},
year = {2017},
doi = {10.1007/s00450-017-0375-2},
url = {}
}
DownloadCorani, G., Benavoli, A., Demšar, J., Mangili, F., Zaffalon, M. (2017). Statistical comparison of classifiers through Bayesian hierarchical modelling. Machine Learning 106(11), pp. 1817–1837.
Statistical comparison of classifiers through Bayesian hierarchical modelling
Authors: Corani, G. and Benavoli, A. and Demšar, J. and Mangili, F. and Zaffalon, M.
Year: 2017
Abstract: Usually one compares the accuracy of two competing classifiers using null hypothesis significance tests (nhst).
Yet the nhst tests suffer from important shortcomings, which
can be overcome by switching to Bayesian hypothesis testing.
We propose a Bayesian hierarchical model that jointly analyzes the cross-validation results obtained by two classifiers on multiple data sets. The model estimates more accurately the difference between classifiers on the individual data sets than the traditional approach of averaging, independently on each data set, the cross-validation results.
It does so by jointly analyzing the results obtained on all data sets, and applying shrinkage to the estimates.
The model eventually returns the posterior probability of the accuracies of the two classifiers being practically equivalent or significantly different.
Published in Machine Learning 106(11), pp. 1817–1837.
Statistical comparison of classifiers through Bayesian hierarchical modelling
@ARTICLE{corani2017a,
title = {Statistical comparison of classifiers through {B}ayesian hierarchical modelling},
journal = {Machine Learning},
volume = {106},
author = {Corani, G. and Benavoli, A. and Demšar, J. and Mangili, F. and Zaffalon, M.},
number = {11},
pages = {1817--1837},
year = {2017},
doi = {10.1007/s10994-017-5641-9},
url = {}
}
DownloadCruder, C., Falla, D., Mangili, F., Azzimonti, L., Araújo, L., Williamon, A., Barbero, M. (2017). Profiling the location and extent of musicians' pain using digital pain drawings. PAIN Practice 18(1), 53–66.
Profiling the location and extent of musicians' pain using digital pain drawings
Authors: Cruder, C. and Falla, D. and Mangili, F. and Azzimonti, L. and Araújo, L. and Williamon, A. and Barbero, M.
Year: 2017
Abstract: BACKGROUND AND AIMS:
According to the existing literature, musicians are at risk to experience a range of musculoskeletal painful conditions. Recently, digital technology has been developed to investigate pain location and pain extent. The aim of this study was to describe pain location and pain extent in musicians using a digital method for pain drawing analysis. Additionally, the association between pain drawing (PD) variables and clinical features in musicians with pain were explored.
MATERIALS AND METHODS:
One hundred fifty-eight musicians (90 women and 68 men; age 22.4±3.6 years) were recruited from Swiss and UK conservatoires. Participants were asked to complete a survey including both background musical information and clinical features, the Quick Dash (QD) questionnaire and the digital PDs.
RESULTS:
Of the 158 participants, 126 musicians (79.7%) reported having pain, with higher prevalence in the areas of the neck and shoulders, the lower back and the right arm. The mean of pain extent was 3.1% ±6.5. The mean of QD was larger for musicians showing the presence of pain than for those without pain. Additionally, the results indicated a positive correlation between QD score and pain extent, and there were significant correlations between age and pain intensity, as well as between pain extent and pain intensity.
CONCLUSIONS:
The high prevalence of pain among musicians has been confirmed using a digital PD. In addition, positive correlations between pain extent and upper limb disability has been demonstrated. Our findings highlight the need for effective prevention and treatment strategies for musicians. This article is protected by copyright. All rights reserved.
Published in PAIN Practice 18(1), Wiley, 53–66.
Profiling the location and extent of musicians' pain using digital pain drawings
@ARTICLE{mangili2017a,
title = {Profiling the location and extent of musicians' pain using digital pain drawings},
journal = {{PAIN} Practice},
publisher = {Wiley},
volume = {18},
author = {Cruder, C. and Falla, D. and Mangili, F. and Azzimonti, L. and Ara\'ujo, L. and Williamon, A. and Barbero, M.},
number = {1},
pages = {53--66},
year = {2017},
doi = {10.1111/papr.12581},
url = {}
}
DownloadGorini, F., Azzimonti, L., Delfanti, G., Scarfò, L., Scielzo, C., Bertilaccio, M.T., Ranghetti, P., Gulino, A., Doglioni, C., Napoli, A.D., Capri, M., Franceschi, C., Calligaris-Cappio, F., Ghia, P., Bellone, M., Dellabona, P., Casorati, G., de Lalla, C. (2017). Invariant NKT cells contribute to Chronic Lymphocytic Leukemia surveillance and prognosis. Blood 129(26), pp. 3440-3451.
Invariant NKT cells contribute to Chronic Lymphocytic Leukemia surveillance and prognosis
Authors: Gorini, F. and Azzimonti, L. and Delfanti, G. and Scarfò, L. and Scielzo, C. and Bertilaccio, M.T. and Ranghetti, P. and Gulino, A. and Doglioni, C. and Napoli, A.D. and Capri, M. and Franceschi, C. and Calligaris-Cappio, F. and Ghia, P. and Bellone, M. and Dellabona, P. and Casorati, G. and de Lalla, C.
Year: 2017
Abstract: Chronic lymphocytic leukemia (CLL) is characterized by the expansion of malignant CD5+ B lymphocytes in blood, bone marrow, and lymphoid organs. CD1d-restricted invariant natural killer T (iNKT) cells are innate-like T lymphocytes strongly implicated in tumor surveillance. We investigated the impact of iNKT cells in the natural history of the disease in the Eμ-Tcl1 (Tcl1) CLL mouse model and 68 CLL patients. We found that Tcl1-CLL cells express CD1d and that iNKT cells critically delay disease onset but become functionally impaired upon disease progression. In patients, disease progression correlates with high CD1d expression on CLL cells and impaired iNKT cells. Conversely, disease stability correlates with negative or low CD1d expression on CLL cells and normal iNKT cells, suggesting indirect leukemia control. iNKT cells indeed hinder CLL survival in vitro by restraining CD1d-expressing nurse-like cells, a relevant proleukemia macrophage population. Multivariable analysis identified iNKT cell frequency as an independent predictor of disease progression. Together, these results support the contribution of iNKT cells to CLL immune surveillance and highlight iNKT cell frequency as a prognostic marker for disease progression.
Published in Blood 129(26), pp. 3440-3451.
Invariant NKT cells contribute to Chronic Lymphocytic Leukemia surveillance and prognosis
@ARTICLE{azzimonti2017a,
title = {Invariant {NKT} cells contribute to {C}hronic {L}ymphocytic {L}eukemia surveillance and prognosis},
journal = {Blood},
volume = {129},
author = {Gorini, F. and Azzimonti, L. and Delfanti, G. and Scarf\`o, L. and Scielzo, C. and Bertilaccio, M.T. and Ranghetti, P. and Gulino, A. and Doglioni, C. and Napoli, A.D. and Capri, M. and Franceschi, C. and Calligaris-Cappio, F. and Ghia, P. and Bellone, M. and Dellabona, P. and Casorati, G. and de Lalla, C.},
number = {26},
pages = {3440-3451},
year = {2017},
doi = {10.1182/blood-2016-11-751065},
url = {}
}
DownloadMangili, F., Bonesana, C., Antonucci, A. (2017). Reliable knowledge-based adaptive tests by credal networks. In Antonucci, A., Cholvy, L., Papini, O. (Eds), Symbolic and Quantitative Approaches to Reasoning with Uncertainty. ECSQARU 2017, Lecture Notes in Computer Science 10369, Springer, Cham, pp. 282–291.
Reliable knowledge-based adaptive tests by credal networks
Authors: Mangili, F. and Bonesana, C. and Antonucci, A.
Year: 2017
Abstract: An adaptive test is a computer-based testing technique which adjusts the sequence of questions on the basis of the estimated ability level of the test taker. We suggest the use of credal networks, a generalization of Bayesian networks based on sets of probability mass functions, to implement adaptive tests exploiting the knowledge of the test developer instead of training on databases of answers. Compared to Bayesian networks, these models might offer higher expressiveness and hence a more reliable modeling of the qualitative expert knowledge. The counterpart is a less straightforward identification of the information-theoretic measure controlling the question-selection and the test-stopping criteria. We elaborate on these issues and propose a sound and computationally feasible procedure. Validation against a Bayesian-network approach on a benchmark about German language proficiency assessments suggests that credal networks can be reliable in assessing the student level and effective in reducing the number of questions required to do it.
Published in Antonucci, A., Cholvy, L., Papini, O. (Eds), Symbolic and Quantitative Approaches to Reasoning with Uncertainty. ECSQARU 2017, Lecture Notes in Computer Science 10369, Springer, Cham, pp. 282–291.
Reliable knowledge-based adaptive tests by credal networks
@INPROCEEDINGS{mangili2017b,
title = {Reliable knowledge-based adaptive tests by credal networks},
editor = {Antonucci, A. and Cholvy, L. and Papini, O.},
publisher = {Springer, Cham},
series = {Lecture Notes in Computer Science},
volume = {10369},
booktitle = {Symbolic and Quantitative Approaches to Reasoning {w}ith Uncertainty. {ECSQARU} 2017 },
author = {Mangili, F. and Bonesana, C. and Antonucci, A.},
pages = {282--291},
year = {2017},
doi = {10.1007/978-3-319-61581-3_26},
url = {}
}
DownloadPiga, D., Benavoli, A. (2017). A unified framework for deterministic and probabilistic d-stability analysis of uncertain polynomial matrices. IEEE Transactions on Automatic Control PP(99).
A unified framework for deterministic and probabilistic d-stability analysis of uncertain polynomial matrices
Authors: Piga, D. and Benavoli, A.
Year: 2017
Abstract: In control theory, we are often interested in robust D-stability analysis, which aims at verifying if all the eigenvalues of an uncertain matrix lie in a given region D. Although many algorithms have been developed to provide conditions for an uncertain matrix to be robustly D-stable, the problem of computing the probability of an uncertain matrix to be D-stable is still unexplored. The goal of this paper is to fill this gap in two directions. First, the only constraint on the stability region D that we impose is that its complement is a semialgebraic set. This comprises many important cases in robust control theory. Second, the D-stability analysis problem is formulated in a probabilistic framework, by assuming that only few probabilistic information is available on the uncertain parameters, such as support and some moments. We will show how to compute the minimum probability that the matrix is D-stable by using convex relaxations based on the theory of moments.
Published in IEEE Transactions on Automatic Control PP(99).
A unified framework for deterministic and probabilistic d-stability analysis of uncertain polynomial matrices
@ARTICLE{piga2017a,
title = {A unified framework for deterministic and probabilistic d-stability analysis of uncertain polynomial matrices},
journal = {{IEEE} Transactions on Automatic Control},
volume = {PP},
author = {Piga, D. and Benavoli, A.},
number = {99},
year = {2017},
doi = {10.1109/TAC.2017.2699281},
url = {}
}
DownloadScanagatta, M., Corani, G., Zaffalon, M. (2017). Improved local search in Bayesian networks structure learning. In Antti Hyttinen, Joe Suzuki, Brandon Malone (Eds), Proceedings of The 3rd International Workshop on Advanced Methodologies for Bayesian Networks (AMBN), Proceedings of Machine Learning Research 73, PMLR, pp. 45-56.
Improved local search in Bayesian networks structure learning
Authors: Scanagatta, M. and Corani, G. and Zaffalon, M.
Year: 2017
Abstract: We present a novel approach for score-based structure learning of Bayesian network, which couples an existing ordering-based algorithm for structure optimization with a novel operator for exploring the neighborhood of a given order in the space of the orderings. Our approach achieves state-of-the-art performances in data sets containing thousands of variables.
Published in Antti Hyttinen, Joe Suzuki, Brandon Malone (Eds), Proceedings of The 3rd International Workshop on Advanced Methodologies for Bayesian Networks (AMBN), Proceedings of Machine Learning Research 73, PMLR, pp. 45-56.
Improved local search in Bayesian networks structure learning
@INPROCEEDINGS{scanagatta2017,
title = {Improved local search in {B}ayesian networks structure learning},
editor = {Antti Hyttinen and Joe Suzuki and Brandon Malone},
publisher = {PMLR},
series = {Proceedings of Machine Learning Research},
volume = {73},
booktitle = {Proceedings of The 3rd International Workshop on Advanced Methodologies for Bayesian Networks ({AMBN})},
author = {Scanagatta, M. and Corani, G. and Zaffalon, M.},
pages = {45-56},
year = {2017},
doi = {},
url = {http://proceedings.mlr.press/v73/scanagatta17a.html}
}
DownloadSoullard, Y., Antonucci, A., Destercke, S. (2017). Technical gestures recognition by set-valued hidden Markov models with prior knowledge. In Ferraro, M. B., Giordani, P., Vantaggi, B., Gagolewski, M., Gil, M. A., Grzegorzewski, P., Hryniewicz, O. (Eds), Soft Methods for Data Science, Advances in Intelligent Systems and Computing 456, Springer, pp. 455–462.
Technical gestures recognition by set-valued hidden Markov models with prior knowledge
Authors: Soullard, Y. and Antonucci, A. and Destercke, S.
Year: 2017
Abstract: Hidden Markov models are popular tools for gesture recognition. Once the generative processes of gestures have been identified, an observation sequence is usually classified as the gesture having the highest likelihood, thus ignoring possible prior information. In this paper, we consider two potential improvements of such methods: the inclusion of prior information, and the possibility of considering convex sets of probabilities (in the likelihoods and the prior) to infer imprecise, but more reliable, predictions when information is insufficient. We apply the proposed approach to technical gestures, typically characterized by severe class imbalance. By modelling such imbalances as a prior information, we achieve more accurate results, while the imprecise quantification is shown to produce more reliable estimates.
Published in Ferraro, M. B., Giordani, P., Vantaggi, B., Gagolewski, M., Gil, M. A., Grzegorzewski, P., Hryniewicz, O. (Eds), Soft Methods for Data Science, Advances in Intelligent Systems and Computing 456, Springer, pp. 455–462.
Note: SMPS 2016: Proceedings of the 8th international conference on Soft Methods in Probability and Statistics
Technical gestures recognition by set-valued hidden Markov models with prior knowledge
@INCOLLECTION{antonucci2016a,
title = {Technical gestures recognition by set-valued hidden {M}arkov models with prior knowledge},
editor = {Ferraro, M. B. and Giordani, P. and Vantaggi, B. and Gagolewski, M. and Gil, M. A. and Grzegorzewski, P. and Hryniewicz, O.},
publisher = {Springer},
series = {Advances in Intelligent Systems and Computing},
volume = {456},
booktitle = {Soft Methods for Data Science},
author = {Soullard, Y. and Antonucci, A. and Destercke, S.},
pages = {455--462},
year = {2017},
doi = {10.1007/978-3-319-42972-4_56},
url = {}
}
DownloadMiranda, E., Zaffalon, M. (2017). Full conglomerability, continuity and marginal extension. In Ferraro, M. B., Giordani, P., Vantaggi, B., Gagolewski, M., Gil, M. A., Grzegorzewski, P., Hryniewicz, O. (Eds), Soft Methods for Data Science, Advances in Intelligent Systems and Computing 456, Springer, pp. 355-362.
Full conglomerability, continuity and marginal extension
Authors: Miranda, E., Zaffalon, M.
Year: 2017
Abstract: We investigate fully conglomerable coherent lower previsions in the sense of Walley, and some particular cases of interest: envelopes of fully conglomerable linear previsions, envelopes of countably additive linear previsions and fully disintegrable linear previsions. We study the connections with continuity and countable super-additivity, and show that full conglomerability can be characterised in terms of a supremum of marginal extension models.
Accepted in Ferraro, M. B., Giordani, P., Vantaggi, B., Gagolewski, M., Gil, M. A., Grzegorzewski, P., Hryniewicz, O. (Eds), Soft Methods for Data Science, Advances in Intelligent Systems and Computing 456, Springer, pp. 355-362.
Note: SMPS 2016: Proceedings of the 8th international conference on Soft Methods in Probability and Statistics
Full conglomerability, continuity and marginal extension
@INCOLLECTION{zaffalon2017a,
title = {Full conglomerability, continuity and marginal extension},
editor = {Ferraro, M. B. and Giordani, P. and Vantaggi, B. and Gagolewski, M. and Gil, M. A. and Grzegorzewski, P. and Hryniewicz, O.},
publisher = {Springer},
series = {Advances in Intelligent Systems and Computing},
volume = {456},
booktitle = {Soft Methods for Data Science},
author = {Miranda, E., Zaffalon, M.},
pages = {355-362},
year = {2017},
doi = {10.1007/978-3-319-42972-4},
url = {}
}
DownloadMiranda, E., Zaffalon, M. (2017). Full conglomerability. Journal of Statistical Theory and Practice 11(4), pp. 634–669.
Full conglomerability
Authors: Miranda, E., Zaffalon, M.
Year: 2017
Abstract: We do a thorough mathematical study of the notion of full conglomerability, that is, conglomerability with respect to all the partitions of an infinite possibility space, in the sense considered by Peter Walley (1991). We consider both the cases of precise and imprecise probability (sets of probabilities). We establish relations between conglomerability and countable additivity, continuity, super-additivity and marginal extension. Moreover, we discuss the special case where a model is conglomerable with respect to a subset of all the partitions, and try to sort out the different notions of conglomerability present in the literature. We conclude that countable additivity, which is routinely used to impose full conglomerability in the precise case, appears to be the most well-behaved way to do so in the imprecise case as well by taking envelopes of countably additive probabilities. Moreover, we characterise these envelopes by means of a number of necessary and sufficient conditions.
Published in Journal of Statistical Theory and Practice 11(4), pp. 634–669.
Full conglomerability
@ARTICLE{zaffalon2017b,
title = {Full conglomerability},
journal = {Journal of Statistical Theory and Practice},
volume = {11},
author = {Miranda, E., Zaffalon, M.},
number = {4},
pages = {634--669},
year = {2017},
doi = {10.1080/15598608.2017.1295890},
url = {}
}
DownloadZaffalon, M., Miranda, E. (2017). Axiomatising incomplete preferences through sets of desirable gambles. Journal of Artificial Intelligence Research 60, pp. 1057–1126.
Axiomatising incomplete preferences through sets of desirable gambles
Authors: Zaffalon, M. and Miranda, E.
Year: 2017
Abstract: We establish the equivalence of two very general theories: the first is the decision-theoretic formalisation of incomplete preferences based on the mixture independence axiom; the second is the theory of coherent sets of desirable gambles (bounded variables) developed in the context of imprecise probability and extended here to vector-valued gambles. Such an equivalence allows us to analyse the theory of incomplete preferences from the point of view of desirability. Among other things, this leads us to uncover an unexpected and clarifying relation: that the notion of `state independence'---the traditional assumption that we can have separate models for beliefs (probabilities) and values (utilities)---coincides with that of `strong independence' in imprecise probability; this connection leads us also to propose much weaker, and arguably more realistic, notions of state independence. Then we simplify the treatment of complete beliefs and values by putting them on a more equal footing. We study the role of the Archimedean condition---which allows us to actually talk of expected utility---, identify some weaknesses and propose alternatives that solve these. More generally speaking, we show that desirability is a valuable alternative foundation to preferences for decision theory that streamlines and unifies a number of concepts while preserving great generality. In addition, the mentioned equivalence shows for the first time how to extend the theory of desirability to imprecise non-linear utility, thus enabling us to formulate one of the most powerful self-consistent theories of reasoning and decision-making available today.
Published in Journal of Artificial Intelligence Research 60, pp. 1057–1126.
Axiomatising incomplete preferences through sets of desirable gambles
@ARTICLE{zaffalon2017c,
title = {Axiomatising incomplete preferences through sets of desirable gambles},
journal = {Journal of Artificial Intelligence Research},
volume = {60},
author = {Zaffalon, M. and Miranda, E.},
pages = {1057--1126},
year = {2017},
doi = {10.1613/jair.5230},
url = {}
}
Download top2016
Antonucci, A., Corani, G. (2016). The multilabel naive credal classifier. International Journal of Approximate Reasoning 83, pp. 320-336.
The multilabel naive credal classifier
Authors: Antonucci, A. and Corani, G.
Year: 2016
Abstract: A credal classifier for multilabel data is presented. This is obtained as an extension of Zaffalon's naive credal classifier to the case of non-exclusive class labels. The dependence relations among the labels are shaped with a tree topology. The classifier, based on a polynomial-time algorithm to compute whether or not a class label is optimal, returns a compact description of the set of optimal sequences of labels. Extensive experiments on real multilabel data show that the classifier gives more robust predictions than its Bayesian counterpart. In practice, when multiple sequences are returned in output, the Bayesian model is more likely to be inaccurate, while the sequences returned by the credal classifier are more likely to include the correct one.
Published in International Journal of Approximate Reasoning 83, pp. 320-336.
The multilabel naive credal classifier
@ARTICLE{antonucci2016c,
title = {The multilabel naive credal classifier},
journal = {International Journal of Approximate Reasoning},
volume = {83},
author = {Antonucci, A. and Corani, G.},
pages = {320-336},
year = {2016},
doi = {10.1016/j.ijar.2016.10.006},
url = {}
}
DownloadBenavoli, A., de Campos, C.P. (2016). Bayesian dependence tests for continuous, binary and mixed continuous-binary variables. Entropy 18(9), pp. 1-24.
Bayesian dependence tests for continuous, binary and mixed continuous-binary variables
Authors: Benavoli, A. and de Campos, C.P.
Year: 2016
Abstract: Tests for dependence of continuous, discrete and mixed continuous-discrete variables are ubiquitous in science. The goal of this paper is to derive Bayesian alternatives to frequentist null hypothesis significance tests for dependence. In particular, we will present three Bayesian tests for dependence of binary, continuous and mixed variables. These tests are nonparametric and based on the Dirichlet Process, which allows us to use the same prior model for all of them. Therefore, the tests are “consistent” among each other, in the sense that the probabilities that variables are dependent computed with these tests are commensurable across the different types of variables being tested. By means of simulations with artificial data, we show the effectiveness of the new tests.
Published in Entropy 18(9), Multidisciplinary Digital Publishing Institute, pp. 1-24.
Bayesian dependence tests for continuous, binary and mixed continuous-binary variables
@ARTICLE{benavoli2016g,
title = {Bayesian dependence tests for continuous, binary and mixed continuous-binary variables},
journal = {Entropy},
publisher = {Multidisciplinary Digital Publishing Institute},
volume = {18},
author = {Benavoli, A. and de Campos, C.P.},
number = {9},
pages = {1-24},
year = {2016},
doi = {10.3390/e18090326},
url = {http://www.mdpi.com/1099-4300/18/9/326}
}
DownloadBenavoli, A., Corani, G., Mangili, F. (2016). Should we really use post-hoc tests based on mean-ranks?. Journal of Machine Learning Research 17(5), pp. 1–10.
Should we really use post-hoc tests based on mean-ranks?
Authors: Benavoli, A. and Corani, G. and Mangili, F.
Year: 2016
Abstract: The statistical comparison of multiple algorithms over multiple data sets is fundamental in machine learning. This is typically carried out by the Friedman test. When the Friedman test rejects the null hypothesis, multiple comparisons are carried out to establish which are the significant differences among algorithms. The multiple comparisons are usually performed using the mean-ranks test. The aim of this technical note is to discuss the inconsistencies of the mean-ranks post-hoc test with the goal of discouraging its use in machine learning as well as in medicine, psychology, etc.. We show that the outcome of the mean-ranks test depends on the pool of algorithms originally included in the experiment. In other words, the outcome of the comparison between algorithms A and B depends also on the performance of the other algorithms included in the original experiment. This can lead to paradoxical situations. For instance the difference between A and B could be declared significant if the pool comprises algorithms C, D, E and not significant if the pool comprises algorithms F, G, H. To overcome these issues, we suggest instead to perform the multiple comparison using a test whose outcome only depends on the two algorithms being compared, such as the sign-test or the Wilcoxon signed-rank test.
Published in Journal of Machine Learning Research 17(5), pp. 1–10.
Should we really use post-hoc tests based on mean-ranks?
@ARTICLE{benavoli2015c,
title = {Should we really use post-hoc tests based on mean-ranks?},
journal = {Journal of Machine Learning Research},
volume = {17},
author = {Benavoli, A. and Corani, G. and Mangili, F.},
number = {5},
pages = {1--10},
year = {2016},
doi = {},
url = {http://jmlr.org/papers/volume17/benavoli16a/benavoli16a.pdf}
}
DownloadBenavoli, A., Facchini, A., Zaffalon, M. (2016). Quantum mechanics: the Bayesian theory generalized to the space of hermitian matrices. Phys. Rev. A 94, 042106.
Quantum mechanics: the Bayesian theory generalized to the space of hermitian matrices
Authors: Benavoli, A. and Facchini, A. and Zaffalon, M.
Year: 2016
Abstract: We consider the problem of gambling on a quantum experiment and enforce rational behavior by a few rules. These rules yield, in the classical case, the Bayesian theory of probability via duality theorems. In our quantum setting, they yield the Bayesian theory generalized to the space of Hermitian matrices. This very theory is quantum mechanics: in fact, we derive all its four postulates from the generalized Bayesian theory. This implies that quantum mechanics is self-consistent. It also leads us to reinterpret the main operations in quantum mechanics as probability rules: Bayes' rule (measurement), marginalization (partial tracing), independence (tensor product). To say it with a slogan, we obtain that quantum mechanics is the Bayesian theory in the complex numbers.
Published in Phys. Rev. A 94, American Physical Society, 042106.
Quantum mechanics: the Bayesian theory generalized to the space of hermitian matrices
@ARTICLE{benavoli2016d,
title = {Quantum mechanics: the {B}ayesian theory generalized to the space of hermitian matrices},
journal = {Phys. Rev. A},
publisher = {American Physical Society},
volume = {94},
author = {Benavoli, A. and Facchini, A. and Zaffalon, M.},
pages = {042106},
year = {2016},
doi = {10.1103/PhysRevA.94.042106},
url = {http://arxiv.org/abs/1605.08177}
}
DownloadBenavoli, A., Facchini, A., Zaffalon, M. (2016). Quantum rational preferences and desirability. In Proceedings of the 1st International Workshop on “imperfect Decision Makers: Admitting Real-world Rationality”, Nips 2016.
Quantum rational preferences and desirability
Authors: Benavoli, A. and Facchini, A. and Zaffalon, M.
Year: 2016
Abstract: We develop a theory of quantum rational decision making in the tradition of Anscombe and Aumann's axiomatisation of preferences on horse lotteries. It is essentially the Bayesian decision theory generalised to the space of Hermitian matrices. Among other things, this leads us to give a representation theorem showing that quantum complete rational preferences are obtained by means of expected utility considerations.
Published in Proceedings of the 1st International Workshop on “imperfect Decision Makers: Admitting Real-world Rationality”, Nips 2016 ArXiv e-prints 1610.06764.
Quantum rational preferences and desirability
@INPROCEEDINGS{benavoli2016h,
title = {Quantum rational preferences and desirability},
journal = {{ArXiv} {e}-{p}rints 1610.06764},
booktitle = {Proceedings of the 1st International Workshop on ``imperfect Decision Makers: Admitting Real-{w}orld Rationality'', Nips 2016},
author = {Benavoli, A. and Facchini, A. and Zaffalon, M.},
year = {2016},
doi = {},
url = {http://arxiv.org/abs/1610.06764}
}
DownloadBenavoli, A., Piga, D. (2016). A probabilistic interpretation of set-membership filtering: application to polynomial systems through polytopic bounding. Automatica 70, pp. 158–172.
A probabilistic interpretation of set-membership filtering: application to polynomial systems through polytopic bounding
Authors: Benavoli, A. and Piga, D.
Year: 2016
Abstract: Set-membership estimation is usually formulated in the context of set-valued calculus and no probabilistic calculations are necessary. In this paper, we show that set-membership estimation can be equivalently formulated in the probabilistic setting by employing sets of probability measures. Inference in set-membership estimation is thus carried out by computing expectations with respect to the updated set of probability measures P as in the probabilistic case. In particular, it is shown that inference can be performed by solving a particular semi-infinite linear programming problem, which is a special case of the truncated moment problem in which only the zero-th order moment is known (i.e., the support). By writing the dual of the above semi- infinite linear programming problem, it is shown that, if the nonlinearities in the measurement and process equations are polynomial and if the bounding sets for initial state, process and measurement noises are described by polynomial inequalities, then an approximation of this semi-infinite linear programming problem can efficiently be obtained by using the theory of sum- of-squares polynomial optimization. We then derive a smart greedy procedure to compute a polytopic outer-approximation of the true membership-set, by computing the minimum-volume polytope that outer-bounds the set that includes all the means computed with respect to P.
Published in Automatica 70, pp. 158–172.
A probabilistic interpretation of set-membership filtering: application to polynomial systems through polytopic bounding
@ARTICLE{benavoli2016a,
title = {A probabilistic interpretation of set-membership filtering: application to polynomial systems through polytopic bounding},
journal = {Automatica},
volume = {70},
author = {Benavoli, A. and Piga, D.},
pages = {158--172},
year = {2016},
doi = {10.1016/j.automatica.2016.03.021},
url = {}
}
DownloadBucher, D., Cellina, F., Mangili, F., Raubal, M., Rudel, R., Rizzoli, A.E., Elabed, O. (2016). Exploiting fitness apps for sustainable mobility - challenges deploying the GoEco! App. In Proceedings of the 2016 conference ICT for Sustainability, Advances in Computer Science Research, pp. 89 - 98.
Exploiting fitness apps for sustainable mobility - challenges deploying the GoEco! App
Authors: Bucher, D. and Cellina, F. and Mangili, F. and Raubal, M. and Rudel, R. and Rizzoli, A.E. and Elabed, O.
Year: 2016
Abstract: The large interest in analyzing one’s own fitness led
to the development of more and more powerful smartphone applications. Most are capable of tracking a user’s position and mode of locomotion, data that do not only reflect personal health, but also mobility choices. A large field of research is concerned with mobility analysis and planning for a variety of reasons, including sustainable transport. Collecting data on mobility behavior using fitness tracker apps is a tempting choice, because they include many of the desired functions, most people own a smartphone and installing a fitness tracker is quick and convenient. However, as their original focus is on measuring fitness behavior, there are a number of difficulties in their usage for mobility tracking. In this paper we denote the various challenges we faced when deploying GoEco! Tracker (an app using the Moves R fitness tracker to collect mobility measurements), and provide an analysis on how to best overcome them. Finally, we summarize findings after one
month of large scale testing with a few hundred users within the GoEco! living lab performed in Switzerland.
Published in Proceedings of the 2016 conference ICT for Sustainability, Advances in Computer Science Research, pp. 89 - 98.
Exploiting fitness apps for sustainable mobility - challenges deploying the GoEco! App
@INPROCEEDINGS{mangili2016c,
title = {Exploiting fitness apps for sustainable mobility - challenges deploying the {GoEco}! App},
series = {Advances in Computer Science Research},
booktitle = {Proceedings of the 2016 {c}onference {ICT} for Sustainability},
author = {Bucher, D. and Cellina, F. and Mangili, F. and Raubal, M. and Rudel, R. and Rizzoli, A.E. and Elabed, O.},
pages = {89 - 98},
year = {2016},
doi = {doi:10.2991/ict4s-16.2016.11},
url = {}
}
DownloadCabañas, R., Antonucci, A., Cano, A., Gómez-Olmedo, M. (2016). Evaluating interval-valued influence diagrams. International Journal of Approximate Reasoning 80, pp. 393-411.
Evaluating interval-valued influence diagrams
Authors: Cabañas, R. and Antonucci, A. and Cano, A. and Gómez-Olmedo, M.
Year: 2016
Abstract: Influence diagrams are probabilistic graphical models used to represent and solve sequential decision problems under uncertainty. Sharp numerical values are required to quantify probabilities and utilities. This might be an issue with real models, whose parameters are typically obtained from expert judgements or partially reliable data. We consider an interval-valued quantification of the parameters to gain realism in the modeling and evaluate the sensitivity of the inferences with respect to perturbations in the sharp values of the parameters. An extension of the classical influence diagrams formalism to support such interval-valued potentials is presented. The variable elimination and arc reversal inference algorithms are generalized to cope with these models. At the price of an outer approximation, the extension keeps the same complexity as with sharp values. Numerical experiments show improved performances with respect to previous methods. As a natural application, we propose these models for practical sensitivity analysis in traditional influence diagrams. The maximum perturbation level on single or multiple parameters preserving the optimal strategy can be computed. This allows the identification of the parameters deserving a more careful elicitation.
Published in International Journal of Approximate Reasoning 80, pp. 393-411.
Evaluating interval-valued influence diagrams
@ARTICLE{antonucci2016b,
title = {Evaluating interval-valued influence diagrams},
journal = {International Journal of Approximate Reasoning},
volume = {80},
author = {Caba\~nas, R. and Antonucci, A. and Cano, A. and G\'omez-Olmedo, M.},
pages = {393-411},
year = {2016},
doi = {10.1016/j.ijar.2016.05.004},
url = {}
}
Downloadde Campos, C.P., Benavoli, A. (2016). Joint analysis of multiple algorithms and performance measures. New Generation Computing, pp. 1–18.
Joint analysis of multiple algorithms and performance measures
Authors: de Campos, C.P. and Benavoli, A.
Year: 2016
Abstract: There has been an increasing interest in the development of new methods using Pareto optimality to deal with multi-objective criteria (for example, accuracy and time complexity). Once one has developed an approach to a problem of interest, the problem is then how to compare it with the state of art. In machine learning, algorithms are typically evaluated by comparing their performance on different data sets by means of statistical tests. Standard tests used for this purpose are able to consider jointly neither performance measures nor multiple competitors at once. The aim of this paper is to resolve these issues by developing statistical procedures that are able to account for multiple competing measures at the same time and to compare multiple algorithms altogether. In particular, we develop two tests: a frequentist procedure based on the generalized likelihood ratio test and a Bayesian procedure based on a multinomial-Dirichlet conjugate model. We further extend them by discovering conditional independences among measures to reduce the number of parameters of such models, as usually the number of studied cases is very reduced in such comparisons. Data from a comparison among general purpose classifiers are used to show a practical application of our tests.
Published in New Generation Computing, pp. 1–18.
Joint analysis of multiple algorithms and performance measures
@ARTICLE{deCampos2016,
title = {Joint analysis of multiple algorithms and performance measures},
journal = {New Generation Computing},
author = {de Campos, C.P. and Benavoli, A.},
pages = {1--18},
year = {2016},
doi = {10.1007/s00354-016-0005-8},
url = {http://people.idsia.ch/~alessio/decampos-benavoli-ngc2016.pdf}
}
Downloadde Campos, C.P., Corani, G., Scanagatta, M., Cuccu, M., Zaffalon, M. (2016). Learning extended tree augmented naive structures. International Journal of Approximate Reasoning. 68, pp. 153–163.
Learning extended tree augmented naive structures
Authors: de Campos, C.P. and Corani, G. and Scanagatta, M. and Cuccu, M. and Zaffalon, M.
Year: 2016
Abstract: This work proposes an extended version of the well-known tree-augmented naive Bayes (TAN) classifier where the structure learning step is performed without requiring features to be connected to the class. Based on a modification of Edmonds' algorithm, our structure learning procedure explores a superset of the structures that are considered by TAN, yet achieves global optimality of the learning score function in a very efficient way (quadratic in the number of features, the same complexity as learning TANs). We enhance our procedure with a new score function that only takes into account arcs that are relevant to predict the class, as well as an optimization over the equivalent sample size during learning. These ideas may be useful for structure learning of Bayesian networks in general. A range of experiments show that we obtain models with better prediction accuracy than Naive Bayes and TAN, and comparable to the accuracy of the state-of-the-art classifier averaged one-dependence estimator (AODE). We release our implementation of ETAN so that it can be easily installed and run within Weka.
Accepted in International Journal of Approximate Reasoning. 68, pp. 153–163.
Learning extended tree augmented naive structures
@ARTICLE{decampos2015a,
title = {Learning extended tree augmented naive structures},
journal = {International Journal of Approximate Reasoning.},
volume = {68},
author = {de Campos, C.P. and Corani, G. and Scanagatta, M. and Cuccu, M. and Zaffalon, M.},
pages = {153--163},
year = {2016},
doi = {10.1016/j.ijar.2015.04.006},
url = {}
}
DownloadCorani, G., Scanagatta, M. (2016). Air pollution prediction via multi-label classification. Environmental Modelling & Software 80, pp. 259–264.
Air pollution prediction via multi-label classification
Authors: Corani, G. and Scanagatta, M.
Year: 2016
Abstract: A Bayesian network classifier can be used to estimate the probability of an air pollutant overcoming a certain threshold. Yet multiple predictions are typically required regarding variables which are stochastically dependent, such as ozone measured in multiple stations or assessed according to by different indicators. The common practice (independent approach) is to devise an independent classifier for each class variable being predicted; yet this approach overlooks the dependencies among the class variables. By appropriately modeling such dependencies one can improve the accuracy of the forecasts. We address this problem by designing a multi-label classifier, which simultaneously predict multiple air pollution variables. To this end we design a multi-label classifier based on Bayesian networks and learn its structure through structural learning. We present experiments in three different case studies regarding the prediction of PM2.5 and ozone. The multi-label classifier outperforms the independent approach, allowing to take better decisions.
Published in Environmental Modelling & Software 80, pp. 259–264.
Air pollution prediction via multi-label classification
@ARTICLE{corani2016a,
title = {Air pollution prediction via multi-label classification},
journal = {Environmental Modelling & Software},
volume = {80},
author = {Corani, G. and Scanagatta, M.},
pages = {259--264},
year = {2016},
doi = {10.1016/j.envsoft.2016.02.030},
url = {}
}
DownloadFu, S. (2016). Hierarchical Bayesian LASSO for a negative binomial regression. Journal of Statistical Computation and Simulation.
Hierarchical Bayesian LASSO for a negative binomial regression
Authors: Fu, S.
Year: 2016
Abstract: Numerous researches have been carried out to explain the relationship between the count data y and numbers of covariates x through a generalized linear model (GLM). This paper proposes a hierarchical Bayesian least absolute shrinkage and selection operator (LASSO) solution using six different prior models to the negative binomial regression. Latent variables Z have been introduced to simplify the GLM to a standard linear regression model. The proposed models regard two conjugate zero-mean Normal priors for the regression parameters and three independent priors for the variance: the Exponential, Inverse-Gamma and Scaled Inverse- distributions. Different types of priors result in different amounts of shrinkage. A Metropolis–Hastings-within-Gibbs algorithm is used to compute the posterior distribution of the parameters of interest through a data augmentation process. Based on the posterior samples, an original double likelihood ratio test statistic have been proposed to choose the most relevant covariates and shrink the insignificant coefficients to zero. Numerical experiments on a real-life data set prove that Bayesian LASSO methods achieved significantly better predictive accuracy and robustness than the classical maximum likelihood estimation and the standard Bayesian inference.
Published in Journal of Statistical Computation and Simulation.
Hierarchical Bayesian LASSO for a negative binomial regression
@ARTICLE{shuaiFu2015a,
title = {Hierarchical {B}ayesian {LASSO} for a negative binomial regression},
journal = {Journal of Statistical Computation and Simulation},
author = {Fu, S.},
year = {2016},
doi = {10.1080/00949655.2015.1106541},
url = {}
}
DownloadGuerciotti, B., Vergara, C., Azzimonti, L., Forzenigo, L., Buora, A., Biondetti, P., Domanin, M. (2016). Computational study of the fluid-dynamics in carotids before and after endarterectomy. Journal of Biomechanics 49(1), pp. 26–38.
Computational study of the fluid-dynamics in carotids before and after endarterectomy
Authors: Guerciotti, B. and Vergara, C. and Azzimonti, L. and Forzenigo, L. and Buora, A. and Biondetti, P. and Domanin, M.
Year: 2016
Abstract: In this work, we provide a computational study of the effects of carotid endarterectomy (CEA) on the fluid-dynamics at internal carotid bifurcations. We perform numerical simulations in real geometries of the same patients before and after CEA, using patient-specific boundary data obtained by Echo-Color Doppler measurements. We analyze four patients with a primary closure and other four where a patch was used to close arteriotomies. The results show that (i) CEA is able to restore physiological fluid-dynamic conditions; (ii) among the post-operative cases, the presence of patch leads to local hemodynamic conditions which might imply a higher risk of restenosis in comparison with the cases without patch.
Published in Elsevier (Ed), Journal of Biomechanics 49(1), pp. 26–38.
Computational study of the fluid-dynamics in carotids before and after endarterectomy
@ARTICLE{azzimonti2016a,
title = {Computational study of the fluid-dynamics in carotids before and after endarterectomy},
journal = {Journal of Biomechanics},
editor = {Elsevier},
volume = {49},
author = {Guerciotti, B. and Vergara, C. and Azzimonti, L. and Forzenigo, L. and Buora, A. and Biondetti, P. and Domanin, M.},
number = {1},
pages = {26--38},
year = {2016},
doi = {10.1016/j.jbiomech.2015.11.009},
url = {}
}
DownloadMangili, F. (2016). A prior near-ignorance Gaussian process model for nonparametric regression. International Journal of Approximate Reasoning.
A prior near-ignorance Gaussian process model for nonparametric regression
Authors: Mangili, F.
Year: 2016
Abstract: Abstract This paper proposes a prior near-ignorance model for regression based on a set of Gaussian Processes (GP). GPs are natural prior distributions for Bayesian regression. They offer a great modeling flexibility and have found widespread application in many regression problems. However, a GP requires the prior elicitation of its mean function, which represents our prior belief about the shape of the regression function, and of the covariance between any two function values. In the absence of prior information, it may be difficult to fully specify these infinite dimensional parameters. In this work, by modeling the prior mean of the GP as a linear combination of a set of basis functions and assuming as prior for the combination coefficients a set of conjugate distributions obtained as limits of truncate exponential priors, we have been able to model prior ignorance about the mean of the GP. The resulting model satisfies translation invariance, learning and, under some constraints, convergence, which are desirable properties for a prior near-ignorance model. Moreover, it is shown in this paper how this model can be extended to allow for a weaker specification of the GP covariance between function values, by letting each basis function to vary in a set of functions. Application to hypothesis testing has shown how the use of this model induces the capability of automatically detecting when a reliable decision cannot be made based on the available data.
Published in International Journal of Approximate Reasoning.
A prior near-ignorance Gaussian process model for nonparametric regression
@ARTICLE{mangili2016b,
title = {A prior near-ignorance {G}aussian process model for nonparametric regression},
journal = {International Journal of Approximate Reasoning },
author = {Mangili, F.},
year = {2016},
doi = {http://dx.doi.org/10.1016/j.ijar.2016.07.005},
url = {}
}
DownloadMangili, F., Bonesana, C., Antonucci, A., Zaffalon, M., Rubegni, E., Addimando, L. (2016). Adaptive testing by Bayesian networks with application to language assessment. In Micarelli, Alessandro, Stamper, John, Panourgia, Kitty (Eds), Intelligent Tutoring Systems: 13th International Conference, ITS 2016, Zagreb, Croatia, June 7-10, 2016. Proceedings, Lecture Notes in Computer Science, pp. 471–472.
Adaptive testing by Bayesian networks with application to language assessment
Authors: Mangili, F. and Bonesana, C. and Antonucci, A. and Zaffalon, M. and Rubegni, E. and Addimando, L.
Year: 2016
Abstract: We present a general procedure for computerized adaptive testing based on probabilistic graphical models, and show on a real-world benchmark how this procedure can increase the internal consistency of the test and reduce the number of questions without affecting accuracy.
Published in Micarelli, Alessandro, Stamper, John, Panourgia, Kitty (Eds), Intelligent Tutoring Systems: 13th International Conference, ITS 2016, Zagreb, Croatia, June 7-10, 2016. Proceedings, Lecture Notes in Computer Science, pp. 471–472.
Adaptive testing by Bayesian networks with application to language assessment
@INPROCEEDINGS{mangili2016a,
title = {Adaptive testing by {B}ayesian networks with application to language assessment},
editor = {Micarelli, Alessandro and Stamper, John and Panourgia, Kitty},
series = {Lecture Notes in Computer Science},
booktitle = {Intelligent Tutoring Systems: 13th International Conference, {ITS} 2016, Zagreb, Croatia, June 7-10, 2016. Proceedings},
author = {Mangili, F. and Bonesana, C. and Antonucci, A. and Zaffalon, M. and Rubegni, E. and Addimando, L.},
pages = {471--472},
year = {2016},
doi = {},
url = {http://link.springer.com/content/pdf/bbm%3A978-3-319-39583-8%2F1.pdf}
}
DownloadMauá, D.D., Antonucci, A., de Campos, C.P. (2016). Hidden Markov models with set-valued parameters. Neurocomputing 180, pp. 94–107.
Hidden Markov models with set-valued parameters
Authors: Mauá, D.D. and Antonucci, A. and de Campos, C.P.
Year: 2016
Abstract: Hidden Markov models (HMMs) are widely used probabilistic models of sequential data. As with other probabilistic models, they require the speci- fication of local conditional probability distributions, whose assessment can be too difficult and error-prone, especially when data are scarce or costly to acquire. The imprecise HMM (iHMM) generalizes HMMs by allowing the quantification to be done by sets of, instead of single, probability dis- tributions. iHMMs have the ability to suspend judgment when there is not enough statistical evidence, and can serve as a sensitivity analysis tool for standard non-stationary HMMs. In this paper, we consider iHMMs under the strong independence interpretation, for which we develop efficient infer- ence algorithms to address standard HMM usage such as the computation of likelihoods and most probable explanations, as well as performing filter- ing and predictive inference. Experiments with real data show that iHMMs produce more reliable inferences without compromising the computational efficiency.
Published in Neurocomputing 180, pp. 94–107.
Hidden Markov models with set-valued parameters
@ARTICLE{antonucci2015c,
title = {Hidden {M}arkov models with set-valued parameters},
journal = {Neurocomputing},
volume = {180},
author = {Mau\'a, D.D. and Antonucci, A. and de Campos, C.P.},
pages = {94--107},
year = {2016},
doi = {doi:10.1016/j.neucom.2015.08.095},
url = {}
}
DownloadMiranda, E., Zaffalon, M. (2016). Conformity and independence with coherent lower previsions. International Journal of Approximate Reasoning 78, pp. 125–137.
Conformity and independence with coherent lower previsions
Authors: Miranda, E. and Zaffalon, M.
Year: 2016
Abstract: We define the conformity of marginal and conditional models with a joint model within Walley's theory of coherent lower previsions. Loosely speaking, conformity means that the joint can reproduce the marginal and conditional models we started from. By studying conformity with and without additional assumptions of epistemic irrelevance and independence, we establish connections with a number of prominent models in Walley's theory: the marginal extension, the irrelevant natural extension, the independent natural extension and the strong product.
Published in International Journal of Approximate Reasoning 78, pp. 125–137.
Conformity and independence with coherent lower previsions
@ARTICLE{zaffalon2016c,
title = {Conformity and independence with coherent lower previsions},
journal = {International Journal of Approximate Reasoning},
volume = {78},
author = {Miranda, E. and Zaffalon, M.},
pages = {125--137},
year = {2016},
doi = {10.1016/j.ijar.2016.07.004},
url = {}
}
DownloadRancoita, P.M.V., Zaffalon, M., Zucca, E., Bertoni, F., de Campos, C.P. (2016). Bayesian network data imputation with application to survival tree analysis. Computational Statistics and Data Analysis 93, pp. 373–387.
Bayesian network data imputation with application to survival tree analysis
Authors: Rancoita, P.M.V. and Zaffalon, M. and Zucca, E. and Bertoni, F. and de Campos, C.P.
Year: 2016
Abstract: Retrospective clinical datasets are often characterized by a relatively small sample size and many missing data. In this case, a common way for handling the missingness consists in discarding from the analysis patients with missing covariates, further reducing the sample size. Alternatively, if the mechanism that generated the missing allows, incomplete data can be imputed on the basis of the observed data, avoiding the reduction of the sample size and allowing methods to deal with complete data later on. Moreover, methodologies for data imputation might depend on the particular purpose and might achieve better results by considering specific characteristics of the domain. The problem of missing data treatment is studied in the context of survival tree analysis for the estimation of a prognostic patient stratification. Survival tree methods usually address this problem by using surrogate splits, that is, splitting rules that use other variables yielding similar results to the original ones. Instead, our methodology consists in modeling the dependencies among the clinical variables with a Bayesian network, which is then used to perform data imputation, thus allowing the survival tree to be applied on the completed dataset. The Bayesian network is directly learned from the incomplete data using a structural expectation-maximization (EM) procedure in which the maximization step is performed with an exact anytime method, so that the only source of approximation is due to the EM formulation itself. On both simulated and real data, our proposed methodology usually outperformed several existing methods for data imputation and the imputation so obtained improved the stratification estimated by the survival tree (especially with respect to using surrogate splits).
Published in Computational Statistics and Data Analysis 93, pp. 373–387.
Bayesian network data imputation with application to survival tree analysis
@ARTICLE{zaffalon2015b,
title = {Bayesian network data imputation with application to survival tree analysis},
journal = {Computational Statistics and Data Analysis},
volume = {93},
author = {Rancoita, P.M.V. and Zaffalon, M. and Zucca, E. and Bertoni, F. and de Campos, C.P.},
pages = {373--387},
year = {2016},
doi = {10.1016/j.csda.2014.12.008},
url = {}
}
DownloadScanagatta, M., Corani, G., de Campos, C.P., Zaffalon, M. (2016). Learning treewidth-bounded Bayesian networks with thousands of variables. In Daniel D. Lee, Masashi Sugiyama, Ulrike V. Luxburg, Isabelle Guyon, Roman Garnett (Eds), NIPS 2016: Advances in Neural Information Processing Systems 29.
Learning treewidth-bounded Bayesian networks with thousands of variables
Authors: Scanagatta, M. and Corani, G. and de Campos, C.P. and Zaffalon, M.
Year: 2016
Abstract: We present a method for learning treewidth-bounded Bayesian networks from data sets containing thousands of variables. Bounding the treewidth of a Bayesian network greatly reduces the complexity of inferences. Yet, being a global property of the graph, it considerably increases the difficulty of the learning process. Our novel algorithm accomplishes this task, scaling both to large domains and to large treewidths. Our novel approach consistently outperforms the state of the art on experiments with up to thousands of variables.
Published in Daniel D. Lee, Masashi Sugiyama, Ulrike V. Luxburg, Isabelle Guyon, Roman Garnett (Eds), NIPS 2016: Advances in Neural Information Processing Systems 29 NIPS.
Learning treewidth-bounded Bayesian networks with thousands of variables
@INPROCEEDINGS{scanagatta2016a,
title = {Learning treewidth-bounded {B}ayesian networks with thousands of variables},
journal = {{NIPS}},
editor = {Daniel D. Lee and Masashi Sugiyama and Ulrike V. Luxburg and Isabelle Guyon and Roman Garnett},
booktitle = {{NIPS} 2016: Advances in Neural Information Processing Systems 29},
author = {Scanagatta, M. and Corani, G. and de Campos, C.P. and Zaffalon, M.},
year = {2016},
doi = {},
url = {http://papers.nips.cc/paper/6232-learning-treewidth-bounded-bayesian-networks-with-thousands-of-variables}
}
Download top2015
Antonucci, A., Corani, G. (2015). The multilabel naive credal classifier. In Augustin, T., Doria, S., Miranda, E., Quaeghebeur, E. (Eds), ISIPTA '15: Proceedings of the Ninth International Symposium on Imprecise Probability: Theories and Applications, SIPTA, pp. 27–36.
The multilabel naive credal classifier
Authors: Antonucci, A. and Corani, G.
Year: 2015
Abstract: We present a credal classifier for multilabel data. The model generalizes the naive credal classifier to the multilabel case. An imprecise-probabilistic quantifi- cation is achieved by means of the imprecise Dirichlet model in its global formulation. A polynomial-time algorithm to compute whether or not a label is opti- mal according to the maximality criterion is derived. Experimental results show the importance of robust predictions in multilabel problems.
Published in Augustin, T., Doria, S., Miranda, E., Quaeghebeur, E. (Eds), ISIPTA '15: Proceedings of the Ninth International Symposium on Imprecise Probability: Theories and Applications, SIPTA, pp. 27–36.
The multilabel naive credal classifier
@INPROCEEDINGS{antonucci2015a,
title = {The multilabel naive credal classifier},
editor = {Augustin, T. and Doria, S. and Miranda, E. and Quaeghebeur, E.},
publisher = {SIPTA},
booktitle = {{ISIPTA} '15: Proceedings of the Ninth International Symposium on Imprecise Probability: Theories and Applications},
author = {Antonucci, A. and Corani, G.},
pages = {27--36},
year = {2015},
doi = {},
url = {http://www.sipta.org/isipta15/data/paper/32.pdf}
}
DownloadAntonucci, A., de Rosa, R., Giusti, A., Cuzzolin, F. (2015). Robust classification of multivariate time series by imprecise hidden Markov models. International Journal of Approximate Reasoning 56(B), pp. 249–263.
Robust classification of multivariate time series by imprecise hidden Markov models
Authors: Antonucci, A. and de Rosa, R. and Giusti, A. and Cuzzolin, F.
Year: 2015
Abstract: A novel technique to classify time series with imprecise hidden Markov models is presented. The learning of these models is achieved by coupling the EM algorithm with the imprecise Dirichlet model. In the stationarity limit, each model corresponds to an imprecise mixture of Gaussian densities, this reducing the problem to the classification of static, imprecise-probabilistic, information. Two classifiers, one based on the expected value of the mixture, the other on the Bhattacharyya distance between pairs of mixtures, are developed. The computation of the bounds of these descriptors with respect to the imprecise quantification of the parameters is reduced to, respectively, linear and quadratic optimization tasks, and hence efficiently solved. Classification is performed by extending the k-nearest neighbors approach to interval-valued data. The classifiers are credal, this means that multiple class labels can be returned in the output. Experiments on benchmark datasets for computer vision show that these methods achieve the required robustness whilst outperforming other precise and imprecise methods.
Published in International Journal of Approximate Reasoning 56(B), pp. 249–263.
Robust classification of multivariate time series by imprecise hidden Markov models
@ARTICLE{antonucci2014c,
title = {Robust classification of multivariate time series by imprecise hidden {M}arkov models},
journal = {International Journal of Approximate Reasoning},
volume = {56},
author = {Antonucci, A. and de Rosa, R. and Giusti, A. and Cuzzolin, F.},
number = {B},
pages = {249--263},
year = {2015},
doi = {10.1016/j.ijar.2014.07.005},
url = {}
}
DownloadAntonucci, A., Scanagatta, M., Mauà, D.D., de Campos, C.P. (2015). Early classification of time series by hidden Markov models with set-valued parameters . In Proceedings of the NIPS Time Series Workshop 2015.
Early classification of time series by hidden Markov models with set-valued parameters
Authors: Antonucci, A. and Scanagatta, M. and Mauà, D.D. and de Campos, C.P.
Year: 2015
Abstract: Hidden Markov models are popular tools for the modeling of multivariate time series. A set-valued quantification of the parameters might increase realism in the description of non-stationarity. A recent work shows that the computation of the bounds of the likelihood of a sequence with respect to such imprecise quantifica- tion can be performed in the same polynomial time of a precise computation with sharp values. This is the basis for a credal classifier of time series. For each training sequence we learn a set-valued model and compute the interval likelihood of the test sequence. The returned classes are those of the models associated to the undominated intervals. The approach is particularly accurate on the instances for which single classes are returned. We therefore apply this method to early classi- fication of streaming data. As soon as the credal classifier returns a single output we assign a class label even if the stream is not terminated. Tests on a speech recognition benchmark suggest that the proposed approach might outperform a thresholding of the precise likelihoods with the classical models.
Published in Proceedings of the NIPS Time Series Workshop 2015.
Early classification of time series by hidden Markov models with set-valued parameters
@INPROCEEDINGS{antonucci2015d,
title = {Early classification of time series by hidden {M}arkov models with set-valued parameters },
booktitle = {Proceedings of the {NIPS} Time Series Workshop 2015},
author = {Antonucci, A. and Scanagatta, M. and Mau\`a, D.D. and de Campos, C.P.},
year = {2015},
doi = {},
url = {https://sites.google.com/site/nipsts2015/home}
}
DownloadAzzimonti, L., Sangalli, L.M., Secchi, P., Domanin, M., Nobile, F. (2015). Blood flow velocity field estimation via spatial regression with PDE penalization. Journal of the American Statistical Association, Theory and Methods Section 110(511), pp. 1057–1071.
Blood flow velocity field estimation via spatial regression with PDE penalization
Authors: Azzimonti, L. and Sangalli, L.M. and Secchi, P. and Domanin, M. and Nobile, F.
Year: 2015
Abstract: We propose an innovative method for the accurate estimation of surfaces and spatial fields when prior knowledge of the phenomenon under study is available. The prior knowledge included in the model derives from physics, physiology, or mechanics of the problem at hand, and is formalized in terms of a partial differential equation governing the phenomenon behavior, as well as conditions that the phenomenon has to satisfy at the boundary of the problem domain. The proposed models exploit advanced scientific computing techniques and specifically make use of the finite element method. The estimators have a penalized regression form and the usual inferential tools are derived. Both the pointwise and the areal data frameworks are considered. The driving application concerns the estimation of the blood flow velocity field in a section of a carotid artery, using data provided by echo-color Doppler. This applied problem arises within a research project that aims at studying atherosclerosis pathogenesis. Supplementary materials for this article are available online.
Published in Journal of the American Statistical Association, Theory and Methods Section 110(511), pp. 1057–1071.
Blood flow velocity field estimation via spatial regression with PDE penalization
@ARTICLE{azzimonti2015a,
title = {Blood flow velocity field estimation via spatial regression with {PDE} penalization},
journal = {Journal of the American Statistical Association, Theory and Methods Section},
volume = {110},
author = {Azzimonti, L. and Sangalli, L.M. and Secchi, P. and Domanin, M. and Nobile, F.},
number = {511},
pages = {1057--1071},
year = {2015},
doi = {10.1080/01621459.2014.946036},
url = {}
}
DownloadBenavoli, A., de Campos, C.P. (2015). Statistical tests for joint analysis of performance measures. In Suzuki, J., Ueno, M. (Eds), Advanced Methodologies for Bayesian Networks: Second International Workshop, Ambn 2015, Yokohama, Japan, November 16-18, 2015. Proceedings 9505, Springer International Publishing, Cham, pp. 76–92.
Statistical tests for joint analysis of performance measures
Authors: Benavoli, A. and de Campos, C.P.
Year: 2015
Abstract: Recently there has been an increasing interest in the development of new methods using Pareto optimality to deal with multi-objective criteria (for example, accuracy and architectural complexity). Once one has learned a model based on their devised method, the problem is then how to compare it with the state of art. In machine learning, algorithms are typically evaluated by comparing their performance on different data sets by means of statistical tests. Unfortunately, the standard tests used for this purpose are not able to jointly consider performance measures. The aim of this paper is to resolve this issue by developing statistical procedures that are able to account for multiple competing measures at the same time. In particular, we develop two tests: a frequentist procedure based on the generalized likelihood-ratio test and a Bayesian procedure based on a multinomial-Dirichlet conjugate model. We further extend them by discovering conditional independences among measures to reduce the number of parameter of such models, as usually the number of studied cases is very reduced in such comparisons. Real data from a comparison among general purpose classifiers is used to show a practical application of our tests.
Published in Suzuki, J., Ueno, M. (Eds), Advanced Methodologies for Bayesian Networks: Second International Workshop, Ambn 2015, Yokohama, Japan, November 16-18, 2015. Proceedings 9505, Springer International Publishing, Cham, pp. 76–92.
Statistical tests for joint analysis of performance measures
@INBOOK{Benavoli2015e,
title = {Statistical tests for joint analysis of performance measures},
editor = {Suzuki, J. and Ueno, M.},
publisher = {Springer International Publishing},
address = {Cham},
volume = {9505},
booktitle = {Advanced Methodologies for Bayesian Networks: Second International Workshop, Ambn 2015, Yokohama, Japan, November 16-18, 2015. Proceedings},
author = {Benavoli, A. and de Campos, C.P.},
pages = {76--92},
year = {2015},
doi = {10.1007/978-3-319-28379-1_6},
url = {}
}
DownloadBenavoli, A., Corani, G., Mangili, F., Zaffalon, M. (2015). A Bayesian nonparametric procedure for comparing algorithms. In Francis Bach, David Blei (Eds), Proceedings of the 32th International Conference on Machine Learning (ICML 2015), pp. 1264–1272.
A Bayesian nonparametric procedure for comparing algorithms
Authors: Benavoli, A. and Corani, G. and Mangili, F. and Zaffalon, M.
Year: 2015
Abstract: A fundamental task in machine learning is to compare the performance of multiple algorithms. This is usually performed by the frequentist Friedman test followed by multiple comparisons. This implies dealing with the well-known short- comings of null hypothesis significance tests. We propose a Bayesian approach to overcome these problems. We provide three main contributions. First, we propose a nonparametric Bayesian ver- sion of the Friedman test using a Dirichlet pro- cess (DP) based prior. We show that, from a Bayesian perspective, the Friedman test is an in- ference for a multivariate mean based on an ellip- soid inclusion test. Second, we derive a joint pro- cedure for the multiple comparisons which ac- counts for their dependencies and which is based on the posterior probability computed through the DP. The proposed approach allows verifying the null hypothesis, not only rejecting it. Third, as a practical application we show the results in our algorithm for racing, i.e. identifying the best algorithm among a large set of candidates se- quentially assessed. Our approach consistently outperforms its frequentist counterpart.
Published in Francis Bach, David Blei (Eds), Proceedings of the 32th International Conference on Machine Learning (ICML 2015), pp. 1264–1272.
A Bayesian nonparametric procedure for comparing algorithms
@INPROCEEDINGS{benavoli2015d,
title = {A {B}ayesian nonparametric procedure for comparing algorithms},
editor = { Francis Bach and David Blei},
booktitle = {Proceedings of the 32th International Conference on Machine Learning ({ICML} 2015)},
author = {Benavoli, A. and Corani, G. and Mangili, F. and Zaffalon, M.},
pages = {1264--1272},
year = {2015},
doi = {},
url = {}
}
DownloadBenavoli, A., Mangili, F. (2015). Gaussian processes for Bayesian hypothesis tests on regression functions. In Proceedings of the 18th International Conference on Artificial Intelligence (AISTAT 2015), pp. 74–82.
Gaussian processes for Bayesian hypothesis tests on regression functions
Authors: Benavoli, A. and Mangili, F.
Year: 2015
Abstract: Gaussian processes have been used in different application domains such as classification, regression etc. In this paper we show that they can also be employed as a universal tool for developing a large variety of Bayesian statistical hypothesis tests for regression functions. In particular, we will use GPs for testing whether (i) two functions are equal; (ii) a function is monotone (even accounting for seasonality effects); (iii) a function is periodic; (iv) two functions are proportional. By simulation studies, we will show that, beside being more flexible, GP tests are also competitive in terms of performance with state-of-art algorithms.
Published in Proceedings of the 18th International Conference on Artificial Intelligence (AISTAT 2015), pp. 74–82.
Gaussian processes for Bayesian hypothesis tests on regression functions
@INPROCEEDINGS{benavoli2015a,
title = {Gaussian processes for {B}ayesian hypothesis tests on regression functions},
booktitle = {Proceedings of the 18th International Conference on Artificial Intelligence ({AISTAT} 2015)},
author = {Benavoli, A. and Mangili, F.},
pages = {74--82},
year = {2015},
doi = {},
url = {http://proceedings.mlr.press/v38/benavoli15.html}
}
DownloadBenavoli, A., Mangili, F., Ruggeri, F., Zaffalon, M. (2015). Imprecise Dirichlet process with application to the hypothesis test on the probability that X≤Y. Journal of Statistical Theory and Practice 9, pp. 658–684.
Imprecise Dirichlet process with application to the hypothesis test on the probability that X≤Y
Authors: Benavoli, A. and Mangili, F. and Ruggeri, F. and Zaffalon, M.
Year: 2015
Abstract: The Dirichlet process (DP) is one of the most popular Bayesian nonparametric models. An open problem with the DP is how to choose its infinite-dimensional parameter (base measure) in case of lack of prior information. In this work we present the Imprecise DP (IDP)—a prior near-ignorance DP-based model that does not require any choice of this probability measure. It consists of a class of DPs obtained by letting the normalized base measure of the DP vary in the set of all probability measures. We discuss the tight connections of this approach with Bayesian robustness and in particular prior near-ignorance modeling via sets of probabilities. We use this model to perform a Bayesian hypothesis test on the probability P(X≤Y). We study the theoretical properties of the IDP test (e.g., asymptotic consistency), and compare it with the frequentist Mann-Whitney-Wilcoxon rank test that is commonly employed as a test on P(X≤Y). In particular we will show that our method is more robust, in the sense that it is able to isolate instances in which the aforementioned test is virtually guessing at random.
Published in Journal of Statistical Theory and Practice 9, pp. 658–684.
Imprecise Dirichlet process with application to the hypothesis test on the probability that X≤Y
@ARTICLE{benavoli2015b,
title = {Imprecise {D}irichlet process with application to the hypothesis test on the probability that {X&le};{Y}},
journal = {Journal of Statistical Theory and Practice},
volume = {9},
author = {Benavoli, A. and Mangili, F. and Ruggeri, F. and Zaffalon, M.},
pages = {658--684},
year = {2015},
doi = {10.1080/15598608.2014.985997},
url = {}
}
DownloadCabañas, R., Antonucci, A., Cano, A., Gómez-Olmedo, M. (2015). Variable elimination for interval-valued influence diagrams. In Destercke, S., Denoeux, T. (Eds), Proceedings of the 13th European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty (ECSQARU 2015), Lecture Notes in Computer Science 9161, pp. 541–551.
Variable elimination for interval-valued influence diagrams
Authors: Cabañas, R. and Antonucci, A. and Cano, A. and Gómez-Olmedo, M.
Year: 2015
Abstract: Influence diagrams are probabilistic graphical models used to represent and solve decision problems under uncertainty. Sharp nu- merical values are required to quantify probabilities and utilities. Yet, real models are based on data streams provided by partially reliable sen- sors or experts. We propose an interval-valued quantification of these parameters to gain realism in the modelling and to analyse the sensitiv- ity of the inferences with respect to perturbations of the sharp values. An extension of the classical influence diagrams formalism to support interval-valued potentials is provided. Moreover, a variable elimination algorithm especially designed for these models is developed and evalu- ated in terms of complexity and empirical performances.
Published in Destercke, S., Denoeux, T. (Eds), Proceedings of the 13th European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty (ECSQARU 2015), Lecture Notes in Computer Science 9161, pp. 541–551.
Variable elimination for interval-valued influence diagrams
@INCOLLECTION{antonucci2015b,
title = {Variable elimination for interval-valued influence diagrams},
editor = {Destercke, S. and Denoeux, T.},
series = {Lecture Notes in Computer Science},
volume = {9161},
booktitle = {Proceedings of the 13th European Conference on Symbolic and Quantitative Approaches to Reasoning {w}ith Uncertainty ({ECSQARU} 2015)},
author = {Caba\~nas, R. and Antonucci, A. and Cano, A. and G\'omez-Olmedo, M.},
pages = {541--551},
year = {2015},
chapter = {Symbolic and Quantitative Approaches to Reasoning with Uncertainty},
doi = {10.1007/978-3-319-20807-7_49},
url = {}
}
Downloadde Campos, C.P., Antonucci, A. (2015). Imprecision in machine learning and AI. In The IEEE Intelligent Informatics Bulletin 16(1), IEEE Computer Society, pp. 20–23.
Imprecision in machine learning and AI
Authors: de Campos, C.P. and Antonucci, A.
Year: 2015
Published in The IEEE Intelligent Informatics Bulletin 16(1), IEEE Computer Society, pp. 20–23.
Imprecision in machine learning and AI
@INCOLLECTION{antonucci2015e,
title = {Imprecision in machine learning and {AI}},
publisher = {IEEE Computer Society},
volume = {16},
booktitle = {The {IEEE} Intelligent Informatics Bulletin},
author = {de Campos, C.P. and Antonucci, A.},
number = {1},
pages = {20--23},
year = {2015},
doi = {},
url = {http://www.comp.hkbu.edu.hk/~cib/2015/Dec/iib_vol16no1.pdf}
}
DownloadCorani, G., Benavoli, A. (2015). A Bayesian approach for comparing cross-validated algorithms on multiple data sets. Machine Learning 100(2), pp. 285–304.
A Bayesian approach for comparing cross-validated algorithms on multiple data sets
Authors: Corani, G. and Benavoli, A.
Year: 2015
Abstract: We present a Bayesian approach for making statistical inference about the accuracy (or any other score) of two competing algorithms which have been assessed via cross-validation on multiple data sets. The approach is constituted by two pieces. The first is a novel correlated Bayesian t-test for the analysis of the cross-validation results on a single data set which accounts for the correlation due to the overlapping training sets. The second piece merges the posterior probabilities computed by the Bayesian correlated t test on the different data sets to make inference on multiple data sets. It does so by adopting a Poisson-binomial model. The inferences on multiple data sets account for the different uncertainty of the cross-validation results on the different data sets. It is the first test able to achieve this goal. It is generally more powerful than the signed-rank test if ten runs of cross-validation are performed, as it is anyway generally recommended.
Accepted in Machine Learning 100(2), pp. 285–304.
A Bayesian approach for comparing cross-validated algorithms on multiple data sets
@ARTICLE{corani2015b,
title = {A {B}ayesian approach for comparing cross-validated algorithms on multiple data sets},
journal = {Machine Learning},
volume = {100},
author = {Corani, G. and Benavoli, A.},
number = {2},
pages = {285--304},
year = {2015},
doi = {10.1007/s10994-015-5486-z},
url = {}
}
DownloadCorani, G., Benavoli, A., Mangili, F., Zaffalon, M. (2015). Bayesian hypothesis testing in machine learning. In Proc. ECML PKDD 2015 (European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases), pp. 199–202.
Bayesian hypothesis testing in machine learning
Authors: Corani, G. and Benavoli, A. and Mangili, F. and Zaffalon, M.
Year: 2015
Abstract: Most hypothesis testing in machine learning is done using the frequentist null-hypothesis significance test, which has severe drawbacks. We review recent Bayesian tests which overcome the drawbacks of the frequentist ones.
Published in Proc. ECML PKDD 2015 (European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases), pp. 199–202.
Bayesian hypothesis testing in machine learning
@INPROCEEDINGS{corani2015c,
title = {Bayesian hypothesis testing in machine learning},
booktitle = {Proc. {ECML} {PKDD} 2015 (European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases)},
author = {Corani, G. and Benavoli, A. and Mangili, F. and Zaffalon, M.},
pages = {199--202},
year = {2015},
doi = {10.1007/978-3-319-23461-8_13},
url = {}
}
DownloadCorani, G., Mignatti, A. (2015). Credal model averaging for classification: representing prior ignorance and expert opinions.. International Journal of Approximate Reasoning 56(B), pp. 264–277.
Credal model averaging for classification: representing prior ignorance and expert opinions.
Authors: Corani, G. and Mignatti, A.
Year: 2015
Abstract: Bayesian model averaging (BMA) is the state of the art approach for overcoming model uncertainty. Yet, especially on small data sets, the results yielded by BMA might be sensitive to the prior over the models. Credal model averaging (CMA) addresses this problem by substituting the single prior over the models by a set of priors (credal set). Such approach solves the problem of how to choose the prior over the models and automates sensitivity analysis. We discuss various CMA algorithms for building an ensemble of logistic regressors characterized by different sets of covariates. We show how CMA can be appropriately tuned to the case in which one is prior-ignorant and to the case in which instead domain knowledge is available. CMA detects prior-dependent instances, namely instances in which a different class is more probable depending on the prior over the models. On such instances CMA suspends the judgment, returning multiple classes. We thoroughly compare different BMA and CMA variants on a real case study, predicting presence of Alpine marmot burrows in an Alpine valley. We find that BMA is almost a random guesser on the instances recognized as prior-dependent by CMA.
Published in International Journal of Approximate Reasoning 56(B), pp. 264–277.
Credal model averaging for classification: representing prior ignorance and expert opinions.
@ARTICLE{corani2014a,
title = {Credal model averaging for classification: representing prior ignorance and expert opinions.},
journal = {International Journal of Approximate Reasoning},
volume = {56},
author = {Corani, G. and Mignatti, A.},
number = {B},
pages = {264--277},
year = {2015},
doi = {10.1016/j.ijar.2014.07.001},
url = {}
}
DownloadCorani, G., Mignatti, A. (2015). Robust Bayesian model averaging for the analysis of presence–absence data. Environmental and Ecological Statistics 22(3), pp. 513–534.
Robust Bayesian model averaging for the analysis of presence–absence data
Authors: Corani, G. and Mignatti, A.
Year: 2015
Abstract: When developing a species distribution model, usually one tests several competing models such as logistic regressions characterized by different sets of covariates. Yet, there is an exponential number of subsets of covariates to choose from. This generates the problem of model uncertainty. Bayesian model averaging (BMA) is a state-of-the-art approach to deal with model uncertainty. BMA weights the inferences of multiple models. However, the results yielded by BMA depend on the prior probability assigned to the models. Credal model averaging (CMA) extends BMA towards robustness. It substitutes the single prior over the models by a set of priors. The CMA inferences (e.g., posterior probability of inclusion of a covariate, coefficient of a covariate, probability of presence) are intervals. The interval shows the sensitivity of the BMA estimate on the prior over the models. CMA detects the prior-dependent instances, namely cases in which the most probable outcome becomes presence or absence depending on the adopted prior over the models. On such prior-dependent instances, BMA behaves almost as a random guesser. The weakness of BMA on the prior-dependent instances is to our knowledge pointed out for the first time in the ecological literature. On the prior-dependent instances CMA avoids random guessing acknowledging undecidability. In this way it stimulates the decision maker to convey further information before taking the decision. We provide thorough experiments on different data sets.
Accepted in Environmental and Ecological Statistics 22(3), pp. 513–534.
Robust Bayesian model averaging for the analysis of presence–absence data
@ARTICLE{corani2015a,
title = {Robust {B}ayesian model averaging for the analysis of presence--absence data},
journal = {Environmental and Ecological Statistics},
volume = {22},
author = {Corani, G. and Mignatti, A.},
number = {3},
pages = {513--534},
year = {2015},
doi = {10.1007/s10651-014-0308-1},
url = {}
}
DownloadFu, S. (2015). A hierarchical Bayesian approach to negative binomial regression. Methods and Applications of Analysis 22(4), pp. 409–428.
A hierarchical Bayesian approach to negative binomial regression
Authors: Fu, S.
Year: 2015
Abstract: There is a growing interest in establishing the relationship between the count data y and numerous covariates x through a generalized linear model (GLM), such as explaining the road crash counts from the geometry and environmental factors. This paper proposes a hierarchical Bayesian method to deal with the negative binomial GLM. The Negative Binomial distribution is preferred for modeling nonnegative overdispersed data. The Bayesian inference is chosen to account for prior expert knowledge on regression coefficients in a small sample size setting and the hierarchical structure allows to consider the dependence among the subsets. A Metropolis-Hastings-within-Gibbs algorithm is used to compute the posterior distribution of the parameters of interest through a data augmentation process. The Bayesian approach highly over-performs the classical maximum likelihood estimation in terms of goodness of fit, especially when the sample size decreases and the model complexity increases. Their respective performances have been examined in both the simulated and real-life case studies.
Published in Methods and Applications of Analysis 22(4), pp. 409–428.
A hierarchical Bayesian approach to negative binomial regression
@ARTICLE{shuaiFu2015b,
title = {A hierarchical {B}ayesian approach to negative binomial regression},
journal = {Methods and Applications of Analysis},
volume = {22},
author = {Fu, S.},
number = {4},
pages = {409--428},
year = {2015},
doi = {10.4310/MAA.2015.v22.n4.a4},
url = {}
}
DownloadMangili, F. (2015). A prior near-ignorance Gaussian Process model for nonparametric regression. In Augustin,T., Doria, S., Miranda, E., Quaeghebeur, E. (Eds), ISIPTA '15: Proceedings of the Ninth International Symposium on Imprecise Probability: Theories and Applications, SIPTA, pp. 187–196.
A prior near-ignorance Gaussian Process model for nonparametric regression
Authors: Mangili, F.
Year: 2015
Abstract: A Gaussian Process (GP) defines a distribution over functions and thus it is a natural prior distribution for learning real-valued functions from a set of noisy data.
GPs offer a great modeling flexibility and have found widespread application in many regression problems.
A GP is fully defined by a mean function that represents our prior belief about the shape of the regression function and a covariance function, relating the function values at different covariates.
In the absence of prior information, one typically assumes a GP with zero mean function.
Therefore, a priori, it is assumed that the regression function is constantly equal to zero.
The aim of this paper is to model a situation of prior near-ignorance about the GP mean function. For this we consider the set of all GPs with fixed covariance function and constrant mean function free to vary from $-\infty$ to $+\infty$.
We apply the model with constant mean function to hypothesis testing; in particular we test the equality of two regression functions and show that the use of a prior near-ignorance model allows the test to automatically detect when a reliable decision cannot be made based on the available data.
Finally, we propose a generalization of this model that allows considering other sets of prior mean functions.
Published in Augustin,T., Doria, S., Miranda, E., Quaeghebeur, E. (Eds), ISIPTA '15: Proceedings of the Ninth International Symposium on Imprecise Probability: Theories and Applications, SIPTA, pp. 187–196.
Note: 20-24 July, Pescara, Italy
A prior near-ignorance Gaussian Process model for nonparametric regression
@INPROCEEDINGS{mangili2015b,
title = {A prior near-ignorance {G}aussian {P}rocess model for nonparametric regression},
editor = {Augustin,T. and Doria, S. and Miranda, E. and Quaeghebeur, E.},
publisher = {SIPTA},
booktitle = {{ISIPTA} '15: Proceedings of the Ninth International Symposium on Imprecise Probability: Theories and Applications},
author = {Mangili, F.},
pages = {187--196},
year = {2015},
doi = {},
url = {http://www.sipta.org/isipta15/data/paper/15.pdf}
}
DownloadMangili, F., Benavoli, A. (2015). New prior near-ignorance models on the simplex. International Journal of Approximate Reasoning 56(Part B), pp. 278–306.
New prior near-ignorance models on the simplex
Authors: Mangili, F. and Benavoli, A.
Year: 2015
Abstract: The aim of this paper is to derive new near-ignorance models on the probabil-
ity simplex, which do not directly involve the Dirichlet distribution and, thus,
are alternative to the Imprecise Dirichlet Model (IDM). We focus our investi-
gation on a particular class of distributions on the simplex which is known as
the class of Normalized Infinitely Divisible (NID) distributions; it includes the
Dirichlet distribution as a particular case. For this class it is possible to derive
general formulae for prior and posterior predictive inferences, by exploiting
the Lévy-Khintchine representation theorem. This allows us to generally char-
acterize the near-ignorance properties of the NID class. After deriving these
general properties, we focus our attention on three members of this class. We
will show that one of these near-ignorance models satisfies the representation
invariance principle and, for a given value of the prior strength, always pro-
vides inferences that encompass those of the IDM. The other two models do
not satisfy this principle, but their imprecision depends linearly or almost lin-
early on the number of observed categories; we argue that this is sometimes
a desirable property for a predictive model.
Published in International Journal of Approximate Reasoning 56(Part B), pp. 278–306.
New prior near-ignorance models on the simplex
@ARTICLE{Mangili2014a,
title = {New prior near-ignorance models on the simplex},
journal = {International Journal of Approximate Reasoning},
volume = {56},
author = {Mangili, F. and Benavoli, A.},
number = {Part B},
pages = {278--306},
year = {2015},
doi = {10.1016/j.ijar.2014.08.005},
url = {}
}
DownloadMangili, F., Benavoli, A., de Campos, C.P., Zaffalon, M. (2015). Reliable survival analysis based on the Dirichlet Process. Biometrical Journal 57(6), pp. 1002–1019.
Reliable survival analysis based on the Dirichlet Process
Authors: Mangili, F. and Benavoli, A. and de Campos, C.P. and Zaffalon, M.
Year: 2015
Abstract: We present a robust Dirichlet process for estimating survival functions from samples with right-censored
data. It adopts a prior near-ignorance approach to avoid almost any assumption about the distribution of the
population lifetimes, as well as the need of eliciting an infinite dimensional parameter (in case of lack of
prior information), as it happens with the usual Dirichlet process prior. We show how such model can be
used to derive robust inferences from right-censored lifetime data. Robustness is due to the identification of
the decisions that are prior-dependent, and can be interpreted as an analysis of sensitivity with respect to the
hypothetical inclusion of fictitious new samples in the data. In particular, we derive a nonparametric estimator
of the survival probability and a hypothesis test about the probability that the lifetime of an individual
from one population is shorter than the lifetime of an individual from another. We evaluate these ideas on
simulated data and on the Australian AIDS survival dataset. The methods are publicly available through an
easy-to-use R package.
Published in Biometrical Journal 57(6), pp. 1002–1019.
Reliable survival analysis based on the Dirichlet Process
@ARTICLE{mangili2015a,
title = {Reliable survival analysis based on the {D}irichlet {P}rocess},
journal = {Biometrical Journal},
volume = {57},
author = {Mangili, F. and Benavoli, A. and de Campos, C.P. and Zaffalon, M.},
number = {6},
pages = {1002--1019},
year = {2015},
doi = {10.1002/bimj.201500062},
url = {}
}
DownloadMiranda, E., Zaffalon, M. (2015). On the problem of computing the conglomerable natural extension. International Journal of Approximate Reasoning 56(A), pp. 1–27.
On the problem of computing the conglomerable natural extension
Authors: Miranda, E. and Zaffalon, M.
Year: 2015
Abstract: Embedding conglomerability as a rationality requirement in
probability was among the aims of Walley's behavioural theory of
coherent lower previsions. However, recent work has shown that this
attempt has only been partly successful. If we focus in particular
on the extension of given assessments to a rational and
conglomerable model (in the least-committal way), we have that the
procedure used in Walley's theory, the natural extension,
provides only an approximation to the model that is actually sought
for: the so-called conglomerable natural extension. In this
paper we consider probabilistic assessments in the form of a
coherent lower prevision P, which is another name for a lower
expectation functional, and make an in-depth mathematical study of
the problem of computing the conglomerable natural extension for
this case: that is, where it is defined as the smallest coherent
lower prevision F ≥ P that is conglomerable, in case it exists. Past work has shown that F can be approximated by an increasing sequence (En)n∈ℕ of coherent lower previsions. We solve an open problem by showing that this sequence can consist of infinitely many distinct elements. Moreover, we give sufficient conditions, of quite broad applicability, to make sure that the point-wise limit of the sequence is F in case P is the lower envelope of finitely many linear previsions. In addition, we study the question of the existence of F and its relationship with the notion of marginal extension.
Published in International Journal of Approximate Reasoning 56(A), pp. 1–27.
On the problem of computing the conglomerable natural extension
@ARTICLE{zaffalon2014b,
title = {On the problem of computing the conglomerable natural extension},
journal = {International Journal of Approximate Reasoning},
volume = {56},
author = {Miranda, E. and Zaffalon, M.},
number = {A},
pages = {1--27},
year = {2015},
doi = {10.1016/j.ijar.2014.09.003},
url = {}
}
DownloadMiranda, E., Zaffalon, M. (2015). Independent products in infinite spaces. Journal of Mathematical Analysis and Applications 425(1), pp. 460–488.
Independent products in infinite spaces
Authors: Miranda, E. and Zaffalon, M.
Year: 2015
Abstract: Probabilistic independence, intended as the mutual irrelevance of given variables, can be solidly founded on a notion of self-consistency of an uncertainty model, in particular when probabilities go imprecise. There is nothing in this approach that prevents it from being adopted in very general setups, and yet it has mostly been detailed for variables taking finitely many values. In this mathematical study, we complement previous research by exploring the extent to which such an approach can be generalised. We focus in particular on the independent products of two variables. We characterise the main notions, including some of factorisation and productivity, in the general case where both spaces can be infinite and show that, however, there are situations---even in the case of precise probability---where no independent product exists. This is not the case as soon as at least one space is finite. We study in depth this case at the frontiers of good-behaviour detailing the relations among the most important notions; we show for instance that being an independent product is equivalent to a
certain productivity condition. Then we step back to the general case: we give conditions for the existence of independent products and study ways to get around its inherent limitations.
Published in Journal of Mathematical Analysis and Applications 425(1), pp. 460–488.
Independent products in infinite spaces
@ARTICLE{zaffalon2015a,
title = {Independent products in infinite spaces},
journal = {Journal of Mathematical Analysis and Applications},
volume = {425},
author = {Miranda, E. and Zaffalon, M.},
number = {1},
pages = {460--488},
year = {2015},
doi = {10.1016/j.jmaa.2014.12.049},
url = {}
}
DownloadMiranda, E., Zaffalon, M. (2015). Conformity and independence with coherent lower previsions. In Augustin,T., Doria, S., Miranda, E., Quaeghebeur, E. (Eds), ISIPTA '15: Proceedings of the Ninth International Symposium on Imprecise Probability: Theories and Applications, SIPTA, pp. 197–206.
Conformity and independence with coherent lower previsions
Authors: Miranda, E. and Zaffalon, M.
Year: 2015
Abstract: We study the conformity of marginal unconditional and conditional models with a joint model under assumptions of epistemic irrelevance and independence, within Walley's theory of coherent lower previsions. By doing so, we make a link with a number of prominent models within this theory: the marginal extension, the irrelevant natural extension, the independent natural extension and the strong product.
Published in Augustin,T., Doria, S., Miranda, E., Quaeghebeur, E. (Eds), ISIPTA '15: Proceedings of the Ninth International Symposium on Imprecise Probability: Theories and Applications, SIPTA, pp. 197–206.
Conformity and independence with coherent lower previsions
@INPROCEEDINGS{zaffalon2015c,
title = {Conformity and independence with coherent lower previsions},
editor = {Augustin,T. and Doria, S. and Miranda, E. and Quaeghebeur, E.},
publisher = {SIPTA},
booktitle = {{ISIPTA} '15: Proceedings of the Ninth International Symposium on Imprecise Probability: Theories and Applications},
author = {Miranda, E. and Zaffalon, M.},
pages = {197--206},
year = {2015},
doi = {},
url = {http://www.sipta.org/isipta15/data/paper/16.pdf}
}
DownloadScanagatta, M., de Campos, C.P., Corani, G., Zaffalon, M. (2015). Learning Bayesian networks with thousands of variables. In Corinna Cortes, Neil D. Lawrence, Daniel D. Lee, Masashi Sugiyama, Roman Garnett (Eds), NIPS 2015: Advances in Neural Information Processing Systems 28.
Learning Bayesian networks with thousands of variables
Authors: Scanagatta, M. and de Campos, C.P. and Corani, G. and Zaffalon, M.
Year: 2015
Abstract: We present a method for learning Bayesian networks from data sets containing
thousands of variables without the need for structure constraints. Our approach
is made of two parts. The first is a novel algorithm that effectively explores the
space of possible parent sets of a node. It guides the exploration towards the
most promising parent sets on the basis of an approximated score function that
is computed in constant time. The second part is an improvement of an existing
ordering-based algorithm for structure optimization. The new algorithm provably
achieves a higher score compared to its original formulation. On very large data
sets containing up to ten thousand nodes, our novel approach consistently outper-
forms the state of the art.
Published in Corinna Cortes, Neil D. Lawrence, Daniel D. Lee, Masashi Sugiyama, Roman Garnett (Eds), NIPS 2015: Advances in Neural Information Processing Systems 28 NIPS.
Learning Bayesian networks with thousands of variables
@INPROCEEDINGS{scanagatta2015a,
title = {Learning {B}ayesian networks with thousands of variables},
journal = {{NIPS}},
editor = {Corinna Cortes and Neil D. Lawrence and Daniel D. Lee and Masashi Sugiyama and Roman Garnett},
booktitle = {{NIPS} 2015: Advances in Neural Information Processing Systems 28},
author = {Scanagatta, M. and de Campos, C.P. and Corani, G. and Zaffalon, M.},
year = {2015},
doi = {},
url = {http://papers.nips.cc/paper/5803-learning-bayesian-networks-with-thousands-of-variables}
}
Download top2014
Antonucci, A., de Campos, C.P., Huber, D., Zaffalon, M. (2014). Approximate credal network updating by linear programming with applications to decision making. International Journal of Approximate Reasoning 58, pp. 25–38.
Approximate credal network updating by linear programming with applications to decision making
Authors: Antonucci, A. and de Campos, C.P. and Huber, D. and Zaffalon, M.
Year: 2014
Abstract: Credal nets are probabilistic graphical models which extend Bayesian nets to cope with sets of distributions. An algorithm for approximate credal network updating is presented. The problem in its general formulation is a multilinear optimization task, which can be linearized by an appropriate rule for fixing all the local models apart from those of a single variable. This simple idea can be iterated and quickly leads to accurate inferences. A transformation is also derived to reduce decision making in credal networks based on the maximality criterion to updating. The decision task is proved to have the same complexity of standard inference, being NPPP-complete for general credal nets and NP-complete for polytrees. Similar results are derived for the E-admissibility criterion. Numerical experiments confirm a good performance of the method.
Published in International Journal of Approximate Reasoning 58, pp. 25–38.
Approximate credal network updating by linear programming with applications to decision making
@ARTICLE{antonucci2014e,
title = {Approximate credal network updating by linear programming with applications to decision making},
journal = {International Journal of Approximate Reasoning},
volume = {58},
author = {Antonucci, A. and de Campos, C.P. and Huber, D. and Zaffalon, M.},
pages = {25--38},
year = {2014},
doi = {10.1016/j.ijar.2014.10.003},
url = {}
}
DownloadAntonucci, A., de Campos, C.P., Zaffalon, M. (2014). Probabilistic graphical models. In Augustin, T., Coolen, F., de Cooman, G., Troffaes, M. (Eds), Introduction to Imprecise Probabilities, Wiley, pp. 207–229.
Probabilistic graphical models
Authors: Antonucci, A. and de Campos, C.P. and Zaffalon, M.
Year: 2014
Abstract: This report presents probabilistic graphical models that are based on imprecise probabilities using a simplified language. In particular, the discussion is focused on credal networks and discrete domains. It describes the building blocks of credal networks, algorithms to perform inference, and discusses on complexity results and related work. The goal is to have an easy-to-follow introduction to the topic.
Published in Augustin, T., Coolen, F., de Cooman, G., Troffaes, M. (Eds), Introduction to Imprecise Probabilities, Wiley, pp. 207–229.
Probabilistic graphical models
@INBOOK{antonucci2014a,
title = {Probabilistic graphical models},
editor = {Augustin, T. and Coolen, F. and de Cooman, G. and Troffaes, M.},
publisher = {Wiley},
booktitle = {Introduction to Imprecise Probabilities},
author = {Antonucci, A. and de Campos, C.P. and Zaffalon, M.},
pages = {207--229},
year = {2014},
chapter = {9},
doi = {10.1002/9781118763117.ch9},
url = {}
}
DownloadAntonucci, A., Karlsson, A., Sundgren, D. (2014). Decision making with hierarchical credal sets. In Laurent, A., Strauss, O., Bouchon-Meunier, B., Yager, R.R. (Eds), Information Processing and Management of Uncertainty in Knowledge-based Systems, Communications in Computer and Information Science 444, Springer, pp. 456–465.
Decision making with hierarchical credal sets
Authors: Antonucci, A. and Karlsson, A. and Sundgren, D.
Year: 2014
Abstract: We elaborate on hierarchical credal sets, which are sets of probability mass functions paired with second-order distributions. A new criterion to make decisions based on these models is proposed. This is achieved by sampling from the set of mass functions and considering the Kullback-Leibler divergence from the weighted center of mass of the set. We evaluate this criterion in a simple classification scenario: the results show performance improvements when compared to a credal classifier where the second-order distribution is not taken into account.
Published in Laurent, A., Strauss, O., Bouchon-Meunier, B., Yager, R.R. (Eds), Information Processing and Management of Uncertainty in Knowledge-based Systems, Communications in Computer and Information Science 444, Springer, pp. 456–465.
Decision making with hierarchical credal sets
@INPROCEEDINGS{antonucci2014b,
title = {Decision making with hierarchical credal sets},
editor = {Laurent, A. and Strauss, O. and Bouchon-Meunier, B. and Yager, R.R.},
publisher = {Springer},
series = {Communications in Computer and Information Science},
volume = {444},
booktitle = {Information Processing and Management of Uncertainty in Knowledge-{b}ased Systems},
author = {Antonucci, A. and Karlsson, A. and Sundgren, D.},
pages = {456--465},
year = {2014},
doi = {10.1007/978-3-319-08852-5_47},
url = {}
}
DownloadAzzimonti, L., Nobile, F., Sangalli, L.M., Secchi, P. (2014). Mixed finite elements for spatial regression with PDE penalization. SIAM/ASA Journal on Uncertainty Quantification 2(1), pp. 305-335.
Mixed finite elements for spatial regression with PDE penalization
Authors: Azzimonti, L. and Nobile, F. and Sangalli, L.M. and Secchi, P.
Year: 2014
Abstract: We study a class of models at the interface between statistics and numerical analysis. Specifically, we consider nonparametric regression models for the estimation of spatial fields from pointwise and noisy observations, which account for problem-specific prior information, described in terms of a partial differential equation governing the phenomenon under study. The prior information is incorporated in the model via a roughness term using a penalized regression framework. We prove the well-posedness of the estimation problem, and we resort to a mixed equal order finite element method for its discretization. Moreover, we prove the well-posedness and the optimal convergence rate of the proposed discretization method. Finally the smoothing technique is extended to the case of areal data, particularly interesting in many applications.
Published in SIAM/ASA Journal on Uncertainty Quantification 2(1), pp. 305-335.
Mixed finite elements for spatial regression with PDE penalization
@ARTICLE{azzimonti2014a,
title = {Mixed finite elements for spatial regression with {PDE} penalization},
journal = {{SIAM/ASA} Journal on Uncertainty Quantification},
volume = {2},
author = {Azzimonti, L. and Nobile, F. and Sangalli, L.M. and Secchi, P.},
number = {1},
pages = {305-335},
year = {2014},
doi = {10.1137/130925426},
url = {}
}
DownloadBenavoli, A. (2014). Belief function and multivalued mapping robustness in statistical estimation. International Journal of Approximate Reasoning 55(1, Part 3), pp. 311–329.
Belief function and multivalued mapping robustness in statistical estimation
Authors: Benavoli, A.
Year: 2014
Abstract: We consider the case in which the available knowledge does not allow to specify a precise probabilistic model for the prior and/or likelihood in statistical estimation. We assume that this imprecision can be represented by belief functions models. Thus, we exploit the mathematical structure of belief functions and their equivalent representation in terms of closed convex sets of probabilities to derive robust posterior inferences using Walley theory of imprecise probabilities. Then, we apply these robust models to practical inference problems and we show the connections of the proposed inference method with interval estimation and statistical inference with missing data.
Published in International Journal of Approximate Reasoning 55(1, Part 3), pp. 311–329.
Belief function and multivalued mapping robustness in statistical estimation
@ARTICLE{benavoli2013a,
title = {Belief function and multivalued mapping robustness in statistical estimation},
journal = {International Journal of Approximate Reasoning},
volume = {55},
author = {Benavoli, A.},
number = {1, Part 3},
pages = {311--329},
year = {2014},
doi = {10.1016/j.ijar.2013.04.014},
url = {}
}
DownloadBenavoli, A., Mangili, F., Corani, G., Zaffalon, M., Ruggeri, F. (2014). A Bayesian Wilcoxon signed-rank test based on the Dirichlet process. In Eric P. Xing, Tony Jebara (Eds), Proceedings of the 31st International Conference on Machine Learning (ICML 2014), pp. 1026–1034.
A Bayesian Wilcoxon signed-rank test based on the Dirichlet process
Authors: Benavoli, A. and Mangili, F. and Corani, G. and Zaffalon, M. and Ruggeri, F.
Year: 2014
Abstract: Bayesian methods are ubiquitous in machine learning.
Nevertheless, the analysis of empirical results is
typically performed by frequentist tests. This implies dealing with
null hypothesis significance tests and p-values, even though the shortcomings of such methods
are well known.
We propose a nonparametric Bayesian version of the Wilcoxon signed-rank test using a Dirichlet process (DP) based prior.
We address in two different ways the problem of how to choose the infinite dimensional parameter that characterizes the DP.
The proposed test has all the traditional strengths of the Bayesian approach; for instance, unlike the frequentist tests, it allows verifying the null hypothesis, not only rejecting it, and taking decisions which minimize the expected loss.
Moreover, one of the solutions proposed to model the infinite-dimensional parameter of the DP
allows isolating instances in which the traditional frequentist test is guessing at random.
We show results dealing with the comparison of two classifiers using real and simulated data.
Published in Eric P. Xing, Tony Jebara (Eds), Proceedings of the 31st International Conference on Machine Learning (ICML 2014), pp. 1026–1034.
A Bayesian Wilcoxon signed-rank test based on the Dirichlet process
@INPROCEEDINGS{benavoli2014a,
title = {A {B}ayesian {W}ilcoxon signed-rank test based on the {D}irichlet process},
editor = { Eric P. Xing and Tony Jebara},
booktitle = {Proceedings of the 31st International Conference on Machine Learning ({ICML} 2014)},
author = {Benavoli, A. and Mangili, F. and Corani, G. and Zaffalon, M. and Ruggeri, F.},
pages = {1026--1034},
year = {2014},
doi = {},
url = {http://proceedings.mlr.press/v32/benavoli14.html}
}
DownloadBenavoli, A., Zaffalon, M. (2014). Prior near ignorance for inferences in the k-parameter exponential family. Statistics 49, pp. 1104–1140.
Prior near ignorance for inferences in the k-parameter exponential family
Authors: Benavoli, A. and Zaffalon, M.
Year: 2014
Abstract: This paper proposes a model of prior ignorance about a multivariate variable based on a set of distributions . In particular, we discuss four minimal properties that a model of prior ignorance should satisfy: invariance, near ignorance, learning and convergence. Near ignorance and invariance ensure that our prior model behaves as a vacuous model with respect to some statistical inferences (e.g. mean, credible intervals, etc.) and some transformation of the parameter space. Learning and convergence ensure that our prior model can learn from data and, in particular, that the influence of on the posterior inferences vanishes with increasing numbers of observations. We show that these four properties can all be satisfied by a set of conjugate priors in the multivariate exponential families if the set includes finitely additive probabilities obtained as limits of truncated exponential functions. The obtained set is a model of prior ignorance with respect to the functions (queries) that are commonly used for statistical inferences and, because of conjugacy, it is tractable and easy to elicit. Applications of the model to some practical statistical problems show the effectiveness of the approach.
Published in Statistics 49, pp. 1104–1140.
Prior near ignorance for inferences in the k-parameter exponential family
@ARTICLE{benavoli2014b,
title = {Prior near ignorance for inferences in the k-parameter exponential family},
journal = {Statistics},
volume = {49},
author = {Benavoli, A. and Zaffalon, M.},
pages = {1104--1140},
year = {2014},
doi = {10.1080/02331888.2014.960869},
url = {}
}
DownloadDe Bock, J., de Campos, C.P., Antonucci, A. (2014). Global sensitivity analysis for MAP inference in graphical models. In Ghahramani, Z., Welling, M., Cortes, C., Lawrence, N.D. , Weinberger, K.Q. (Eds), Advances in Neural Information Processing Systems 27 (NIPS 2014), Curran Associates, Inc., pp. 2690–2698.
Global sensitivity analysis for MAP inference in graphical models
Authors: De Bock, J. and de Campos, C.P. and Antonucci, A.
Year: 2014
Abstract: We study the sensitivity of a MAP configuration of a discrete probabilistic graph- ical model with respect to perturbations of its parameters. These perturbations are global, in the sense that simultaneous perturbations of all the parameters (or any chosen subset of them) are allowed. Our main contribution is an exact algorithm that can check whether the MAP configuration is robust with respect to given per- turbations. Its complexity is essentially the same as that of obtaining the MAP configuration itself, so it can be promptly used with minimal effort. We use our algorithm to identify the largest global perturbation that does not induce a change in the MAP configuration, and we successfully apply this robustness measure in two practical scenarios: the prediction of facial action units with posed images and the classification of multiple real public data sets. A strong correlation between the proposed robustness measure and accuracy is verified in both scenarios.
Published in Ghahramani, Z., Welling, M., Cortes, C., Lawrence, N.D. , Weinberger, K.Q. (Eds), Advances in Neural Information Processing Systems 27 (NIPS 2014), Curran Associates, Inc., pp. 2690–2698.
Global sensitivity analysis for MAP inference in graphical models
@INCOLLECTION{antonucci2014f,
title = {Global sensitivity analysis for {MAP} inference in graphical models},
editor = {Ghahramani, Z. and Welling, M. and Cortes, C. and Lawrence, N.D. and Weinberger, K.Q. },
publisher = {Curran Associates, Inc.},
booktitle = {Advances in Neural Information Processing Systems 27 ({NIPS} 2014)},
author = {De Bock, J. and de Campos, C.P. and Antonucci, A.},
pages = {2690--2698},
year = {2014},
doi = {},
url = {http://papers.nips.cc/paper/5472-global-sensitivity-analysis-for-map-inference-in-graphical-models.pdf}
}
Downloadde Campos, C.P., Cuccu, M., Corani, G., Zaffalon, M. (2014). Extended tree augmented naive classifier. In van der Gaag, L., Feelders, A. (Ed), PGM'14: Proceedings of the Seventh European Workshop on Probabilistic Graphical Models, Lecture Notes in Artificial Intelligence 8754, Springer, pp. 176–189.
Extended tree augmented naive classifier
Authors: de Campos, C.P. and Cuccu, M. and Corani, G. and Zaffalon, M.
Year: 2014
Abstract: This work proposes an extended version of the well-known tree-augmented naive Bayes (TAN) classifier where the structure learning step is performed without requiring features to be connected to the class. Based on a modification of Edmonds' algorithm, our structure learning procedure explores a superset of the structures that are considered by TAN, yet achieves global optimality of the learning score function in a very efficient way (quadratic in the number of features, the same complexity as learning TANs). A range of experiments show that we obtain models with better accuracy than TAN and comparable to the accuracy of the state-of-the-art classifier averaged one-dependence estimator.
Published in van der Gaag, L., Feelders, A. (Ed), PGM'14: Proceedings of the Seventh European Workshop on Probabilistic Graphical Models, Lecture Notes in Artificial Intelligence 8754, Springer, pp. 176–189.
Extended tree augmented naive classifier
@INPROCEEDINGS{decampos2014a,
title = {Extended tree augmented naive classifier},
editor = {van der Gaag, L., Feelders, A.},
publisher = {Springer},
series = {Lecture Notes in Artificial Intelligence},
volume = {8754},
booktitle = {{PGM'14}: Proceedings of the Seventh European Workshop on Probabilistic Graphical Models},
author = {de Campos, C.P. and Cuccu, M. and Corani, G. and Zaffalon, M.},
pages = {176--189},
year = {2014},
doi = {10.1007/978-3-319-11433-0_12},
url = {}
}
DownloadCorani, G., Abellán, J., Masegosa, A., Moral, S., Zaffalon, M. (2014). Classification. In Augustin,T., Coolen,F., de Cooman,G., Troffaes,M. (Eds), Introduction to Imprecise Probabilities, Wiley, pp. 261–285.
Classification
Authors: Corani, G. and Abellán, J. and Masegosa, A. and Moral, S. and Zaffalon, M.
Year: 2014
Abstract: This report presents an introduction to credal classification. The discussion focuses on the naive credal classifier and its extensions as well as on credal classifiers based on classification trees. In addition we discuss the metrics suitable for scoring credal classifiers and present some experiments. The goal is to have an easy-to-follow introduction to the topic.
Published in Augustin,T., Coolen,F., de Cooman,G., Troffaes,M. (Eds), Introduction to Imprecise Probabilities, Wiley, pp. 261–285.
Classification
@INCOLLECTION{corani2013b,
title = {Classification},
editor = {Augustin,T. and Coolen,F. and de Cooman,G. and Troffaes,M.},
publisher = {Wiley},
booktitle = {Introduction to Imprecise Probabilities},
author = {Corani, G. and Abell\'an, J. and Masegosa, A. and Moral, S. and Zaffalon, M.},
pages = {261--285},
year = {2014},
chapter = {10},
doi = {},
url = {http://eu.wiley.com/WileyCDA/WileyTitle/productCd-0470973811.html}
}
DownloadCorani, G., Antonucci, A. (2014). Credal Ensembles of Classifiers. Computational Statistics & Data Analysis 71, pp. 818–831.
Credal Ensembles of Classifiers
Authors: Corani, G. and Antonucci, A.
Year: 2014
Abstract: It is studied how to aggregate the probabilistic predictions generated by different SPODE (Super-Parent-One-Dependence Estimators) classifiers. It is shown that aggregating such predictions via compression-based weights achieves a slight but consistent improvement of performance over previously existing aggregation methods, including Bayesian Model Averaging and simple average (the approach adopted by the AODE algorithm). Then, attention is given to the problem of choosing the prior probability distribution over the models; this is an important issue in any Bayesian ensemble of models. To robustly deal with the choice of the prior, the single prior over the models is substituted by a set of priors over the models (credal set), thus obtaining a credal
ensemble of Bayesian classifiers. The credal ensemble recognizes the prior-dependent instances, namely the instances whose most probable class varies when different prior over the models are considered. When faced with prior-dependent instances, the credal ensemble remains reliable by returning a set of classes rather than a single class. Two
credal ensembles of SPODEs are developed; the first generalizes the Bayesian Model Averaging and the second the compression-based aggregation. Extensive experiments show that the novel ensembles compare favorably to traditional methods for aggregating SPODEs and also to previous credal classifiers.
Published in Computational Statistics & Data Analysis 71, pp. 818–831.
Credal Ensembles of Classifiers
@ARTICLE{corani2012f,
title = {Credal {E}nsembles of {C}lassifiers},
journal = {Computational Statistics & Data Analysis},
volume = {71},
author = {Corani, G. and Antonucci, A.},
pages = {818--831},
year = {2014},
doi = {10.1016/j.csda.2012.11.010},
url = {}
}
DownloadCorani, G., Antonucci, A., Mauá, D., Gabaglio, S. (2014). Trading off Speed and Accuracy in Multilabel Classification. In van der Gaag, L., Feelders, A. (Eds), PGM'14: Proceedings of the Seventh European Workshop on Probabilistic Graphical Models, Lecture Notes in Artificial Intelligence 8754, Springer, pp. 145–159.
Trading off Speed and Accuracy in Multilabel Classification
Authors: Corani, G. and Antonucci, A. and Mauá, D. and Gabaglio, S.
Year: 2014
Abstract: In previous work, we devised an approach for multilabel clas-
sification based on an ensemble of Bayesian networks. It was characterized by an efficient structural learning and by high accuracy. Its shortcoming was the high computational complexity of the MAP inference, necessary to identify the most probable joint configuration of all classes.
In this work, we switch from the ensemble approach to the single model approach. This allows important computational savings. The reduction of inference times is exponential in the difference between the treewidth of the single model and the number of classes. We adopt moreover a more sophisticated approach for the structural learning of the class subgraph.
The proposed single models outperforms alternative approaches for multilabel classification such as binary relevance and ensemble of classifier chains.
Published in van der Gaag, L., Feelders, A. (Eds), PGM'14: Proceedings of the Seventh European Workshop on Probabilistic Graphical Models Proc. 7th European Workshop on Probabilistic Graphical Models (PGM '14), Lecture Notes in Artificial Intelligence 8754, Springer, pp. 145–159.
Trading off Speed and Accuracy in Multilabel Classification
@INPROCEEDINGS{corani2014b,
title = {Trading off {S}peed and {A}ccuracy in {M}ultilabel {C}lassification},
journal = {Proc. 7th European Workshop on Probabilistic Graphical Models ({PGM} '14)},
editor = {van der Gaag, L. and Feelders, A.},
publisher = {Springer},
series = {Lecture Notes in Artificial Intelligence},
volume = {8754},
booktitle = {{PGM'14}: Proceedings of the Seventh European Workshop on Probabilistic Graphical Models},
author = {Corani, G. and Antonucci, A. and Mau\'a, D. and Gabaglio, S.},
pages = {145--159},
year = {2014},
doi = {10.1007/978-3-319-11433-0_10},
url = {}
}
DownloadCozman, F.G., de Campos, C.P. (2014). Kuznetsov independence for interval-valued expectations and sets of probability distributions: Properties and algorithms. International Journal of Approximate Reasoning 55(2), pp. 666–682.
Kuznetsov independence for interval-valued expectations and sets of probability distributions: Properties and algorithms
Authors: Cozman, F.G. and de Campos, C.P.
Year: 2014
Abstract: Kuznetsov independence of variables X and Y means that, for any pair of bounded functions f(X) and g(Y), E[f(X)g(Y)]=E[f(X)] *times* E[g(Y)], where E[.] denotes interval-valued expectation and *times* denotes interval multiplication. We present properties of Kuznetsov independence for several variables, and connect it with other concepts of independence in the literature; in particular we show that strong extensions are always included in sets of probability distributions whose lower and upper expectations satisfy Kuznetsov independence. We introduce an algorithm that computes lower expectations subject to judgments of Kuznetsov independence by mixing column generation techniques with nonlinear programming. Finally, we define a concept of conditional Kuznetsov independence, and study its graphoid properties.
Published in International Journal of Approximate Reasoning 55(2), Elsevier, pp. 666–682.
Note: Appeared online in Sept/2013
Kuznetsov independence for interval-valued expectations and sets of probability distributions: Properties and algorithms
@ARTICLE{cozman2013ijar,
title = {Kuznetsov independence for interval-valued expectations and sets of probability distributions: {P}roperties and algorithms},
journal = {International Journal of Approximate Reasoning},
publisher = {Elsevier},
volume = {55},
author = {Cozman, F.G. and de Campos, C.P.},
number = {2},
pages = {666--682},
year = {2014},
doi = {10.1016/j.ijar.2013.09.013},
url = {}
}
DownloadMauá, D.D., de Campos, C.P., Antonucci, A. (2014). Hidden Markov models with imprecisely specified parameters. In Proceedings of the Brazilian Conference on Intelligent Systems.
Hidden Markov models with imprecisely specified parameters
Authors: Mauá, D.D. and de Campos, C.P. and Antonucci, A.
Year: 2014
Abstract: Hidden Markov models (HMMs) are widely used models for sequential data. As with other probabilistic graphical models, they require the specification of precise probability values, which can be too restrictive for some domains, espe- cially when data are scarce or costly to acquire. We present a generalized version of HMMs, whose quantification can be done by sets of, instead of single, probability distributions. Our models have the ability to suspend judgment when there is not enough statistical evidence, and can serve as a sensitivity analysis tool for standard non-stationary HMMs. Efficient inference algorithms are developed to address standard HMM usage such as the computation of likelihoods and most probable explanations. Experiments with real data show that the use of imprecise probabilities leads to more reliable inferences without compromising efficiency.
Published in Proceedings of the Brazilian Conference on Intelligent Systems.
Hidden Markov models with imprecisely specified parameters
@INPROCEEDINGS{antonucci2014d,
title = {Hidden {M}arkov models with imprecisely specified parameters},
booktitle = {Proceedings of the Brazilian Conference on Intelligent Systems},
author = {Mau\'a, D.D. and de Campos, C.P. and Antonucci, A.},
year = {2014},
doi = {},
url = {}
}
DownloadMauá, D.D., de Campos, C.P., Benavoli, A., Antonucci, A. (2014). Probabilistic inference in credal networks: new complexity results. Journal of Artifical Intelligence Research 50, pp. 603–637.
Probabilistic inference in credal networks: new complexity results
Authors: Mauá, D.D. and de Campos, C.P. and Benavoli, A. and Antonucci, A.
Year: 2014
Abstract: Credal networks are graph-based statistical models whose parameters take values in a
set, instead of being sharply specified as in traditional statistical models (e.g., Bayesian
networks). The computational complexity of inferences on such models depends on the
irrelevance/independence concept adopted. In this paper, we study inferential complexity
under the concepts of epistemic irrelevance and strong independence. We show that infer-
ences under strong independence are NP-hard even in trees with binary variables except
for a single ternary one. We prove that under epistemic irrelevance the polynomial-time
complexity of inferences in credal trees is not likely to extend to more general models
(e.g., singly connected topologies). These results clearly distinguish networks that admit
efficient inferences and those where inferences are most likely hard, and settle several open
questions regarding their computational complexity. We show that these results remain
valid even if we disallow the use of zero probabilities. We also show that the computation
of bounds on the probability of the future state in a hidden Markov model is the same
whether we assume epistemic irrelevance or strong independence, and we prove a similar
result for inference in naive Bayes structures. These inferential equivalences are important
for practitioners, as hidden Markov models and naive Bayes structures are used in real
applications of imprecise probability
Published in Journal of Artifical Intelligence Research 50, pp. 603–637.
Probabilistic inference in credal networks: new complexity results
@ARTICLE{maua14jair,
title = {Probabilistic inference in credal networks: new complexity results},
journal = {Journal of Artifical Intelligence Research},
volume = {50},
author = {Mau\'a, D.D. and de Campos, C.P. and Benavoli, A. and Antonucci, A.},
pages = {603--637},
year = {2014},
doi = {10.1613/jair.4355},
url = {}
}
DownloadPolpo, A., de Campos, C.P., Sinha, D., Lipsitz, S., Lin, J. (2014). Transform both sides model: a parametric approach. Computational Statistics and Data Analysis 71, pp. 903–913.
Transform both sides model: a parametric approach
Authors: Polpo, A. and de Campos, C.P. and Sinha, D. and Lipsitz, S. and Lin, J.
Year: 2014
Abstract: A parametric regression model for right-censored data with a log-linear median regression function and a transformation in both response and regression parts, named parametric Transform-Both-Sides (TBS) model, is presented. The TBS model has a parameter that handles data asymmetry while allowing various different distributions for the error, as long as they are unimodal symmetric distributions centered at zero. The discussion is focused on the estimation procedure with five important error distributions (normal, double-exponential, Student's t, Cauchy and logistic) and presents properties, associated functions (that is, survival and hazard functions) and estimation methods based on maximum likelihood and on the Bayesian paradigm. These procedures are implemented in TBSSurvival, an open-source fully documented R package. The use of the package is illustrated and the performance of the model is analyzed using both simulated and real data sets.
Published in Computational Statistics and Data Analysis 71, Elsevier, pp. 903–913.
Note: Appeared online in Jul/2013
Transform both sides model: a parametric approach
@ARTICLE{decampos2013b,
title = {Transform both sides model: a parametric approach},
journal = {Computational Statistics and Data Analysis},
publisher = {Elsevier},
volume = {71},
author = {Polpo, A. and de Campos, C.P. and Sinha, D. and Lipsitz, S. and Lin, J.},
pages = {903--913},
year = {2014},
doi = {10.1016/j.csda.2013.07.023},
url = {}
}
DownloadScanagatta, M., de Campos, C.P., Zaffalon, M. (2014). Min-BDeu and max-BDeu scores for learning Bayesian networks. In van der Gaag, L., Feelders, A. (Eds), PGM'14: Proceedings of the Seventh European Workshop on Probabilistic Graphical Models, Lecture Notes in Artificial Intelligence 8754, Springer, pp. 426–441.
Min-BDeu and max-BDeu scores for learning Bayesian networks
Authors: Scanagatta, M. and de Campos, C.P. and Zaffalon, M.
Year: 2014
Abstract: This work presents two new score functions based on
the Bayesian Dirichlet equivalent uniform (BDeu) score for learning
Bayesian network structures. They consider the sensitivity of BDeu
to varying parameters of the Dirichlet prior. The scores take on the
most adversary and the most beneficial priors among those within a
contamination set around the symmetric one. We build these scores in
such way that they are decomposable and can be computed
efficiently. Because of that, they can be integrated into any
state-of-the-art structure learning method that explores the space
of directed acyclic graphs and allows decomposable scores.
Empirical results suggest that our scores outperform the standard
BDeu score in terms of the likelihood of unseen data and in terms of
edge discovery with respect to the true network, at least when
the training sample size is small. We discuss the relation between
these new scores and the accuracy of inferred models. Moreover, our
new criteria can be used to identify the amount of data after which
learning is saturated, that is, additional data are of little help
to improve the resulting model.
Published in van der Gaag, L., Feelders, A. (Eds), PGM'14: Proceedings of the Seventh European Workshop on Probabilistic Graphical Models, Lecture Notes in Artificial Intelligence 8754, Springer, pp. 426–441.
Min-BDeu and max-BDeu scores for learning Bayesian networks
@INPROCEEDINGS{scanagatta2014a,
title = {Min-{BDeu} and max-{BDeu} scores for learning {B}ayesian networks},
editor = {van der Gaag, L. and Feelders, A.},
publisher = {Springer},
series = {Lecture Notes in Artificial Intelligence},
volume = {8754},
booktitle = {{PGM'14}: Proceedings of the Seventh European Workshop on Probabilistic Graphical Models},
author = {Scanagatta, M. and de Campos, C.P. and Zaffalon, M.},
pages = {426--441},
year = {2014},
doi = {10.1007/978-3-319-11433-0_28},
url = {}
}
DownloadZaffalon, M., Corani, G. (2014). Comments on "Imprecise probability models for learning multinomial distributions from data. Applications to learning credal networks" by Andrés R. Masegosa and Serafín Moral. International Journal of Approximate Reasoning 55(7), pp. 1597–1600.
Comments on "Imprecise probability models for learning multinomial distributions from data. Applications to learning credal networks" by Andrés R. Masegosa and Serafín Moral
Authors: Zaffalon, M. and Corani, G.
Year: 2014
Abstract: We briefly overview the problem of learning probabilities from data using imprecise probability models that express very weak prior beliefs. Then we comment on the new contributions to this question given in the paper by Masegosa and Moral and provide some insights about the performance of their models in data mining experiments of classification.
Published in International Journal of Approximate Reasoning 55(7), pp. 1597–1600.
Comments on "Imprecise probability models for learning multinomial distributions from data. Applications to learning credal networks" by Andrés R. Masegosa and Serafín Moral
@ARTICLE{zaffalon2014a,
title = {Comments on {"Imprecise} probability models for learning multinomial distributions from data. Applications to learning credal networks" by {A}ndr\'es {R}. Masegosa and {S}eraf\'in {M}oral},
journal = {International Journal of Approximate Reasoning},
volume = {55},
author = {Zaffalon, M. and Corani, G.},
number = {7},
pages = {1597--1600},
year = {2014},
doi = {10.1016/j.ijar.2014.05.001},
url = {}
}
Download top2013
Antonucci, A., de Campos, C.P., Huber, D., Zaffalon, M. (2013). Approximating credal network inferences by linear programming. In van der Gaag, L. C. (Ed), Proceedings of the 12th European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty, Lecture Notes in Artificial Intelligence 7958, Springer, Berlin Heidelberg, pp. 13–25.
Approximating credal network inferences by linear programming
Authors: Antonucci, A. and de Campos, C.P. and Huber, D. and Zaffalon, M.
Year: 2013
Abstract: An algorithm for approximate credal network updating is presented. The problem in its general formulation is a multilinear optimization task, which can be linearized by an appropriate rule for fixing all the local models apart from those of a single variable. This simple idea can be iterated and quickly leads to very accurate inferences. The approach can also be specialized to classification with credal networks based on the maximality criterion. A complexity analysis for both the problem and the algorithm is reported together with numerical experiments, which confirm the good performance of the method. While the inner approximation produced by the algorithm gives rise to a classifier which might return a subset of the optimal class set, preliminary empirical results suggest that the accuracy of the optimal class set is seldom affected by the approximate probabilities.
Published in van der Gaag, L. C. (Ed), Proceedings of the 12th European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty, Lecture Notes in Artificial Intelligence 7958, Springer, Berlin Heidelberg, pp. 13–25.
Approximating credal network inferences by linear programming
@INPROCEEDINGS{antonucci2013a,
title = {Approximating credal network inferences by linear programming},
editor = {van der Gaag, L. C.},
publisher = {Springer},
address = {Berlin Heidelberg},
series = {Lecture Notes in Artificial Intelligence},
volume = {7958},
booktitle = {Proceedings of the 12th European Conference on Symbolic and Quantitative Approaches to Reasoning {w}ith Uncertainty},
author = {Antonucci, A. and de Campos, C.P. and Huber, D. and Zaffalon, M.},
pages = {13--25},
year = {2013},
doi = {10.1007/978-3-642-39091-3_2},
url = {}
}
DownloadAntonucci, A., Corani, G., Mauá, D.D., Gabaglio, S. (2013). An ensemble of Bayesian networks for multilabel classification. In Proceedings of the 23rd International Joint Conference on Artificial Intelligence (IJCAI-13), pp. 1220–1225.
An ensemble of Bayesian networks for multilabel classification
Authors: Antonucci, A. and Corani, G. and Mauá, D.D. and Gabaglio, S.
Year: 2013
Abstract: We present a novel approach for multilabel classification based on an ensemble of Bayesian networks. The class variables are connected by a tree; each model of the ensemble uses a different class as root of the tree. We assume the features to be conditionally independent given the classes, thus generalizing the naive Bayes assumption to the multiclass case. This assumption allows us to optimally
identify the correlations between classes and features; such correlations are moreover shared across all models of the ensemble. Inferences are drawn from the ensemble via logarithmic opinion pooling. To minimize Hamming loss, we compute the marginal probability of the classes by running standard inference on each Bayesian network in the ensemble, and then pooling the inferences. To instead minimize the subset 0/1 loss, we pool the joint distributions of each model and cast the problem as a MAP inference in the corresponding graphical
model. Experiments show that the approach is competitive with state-of-the-art methods for multilabel classification.
Published in Proceedings of the 23rd International Joint Conference on Artificial Intelligence (IJCAI-13), pp. 1220–1225.
An ensemble of Bayesian networks for multilabel classification
@INPROCEEDINGS{antonucci2013d,
title = {An ensemble of {B}ayesian networks for multilabel classification},
booktitle = {Proceedings of the 23rd International Joint Conference on Artificial Intelligence ({IJCAI}-13)},
author = {Antonucci, A. and Corani, G. and Mau\'a, D.D. and Gabaglio, S.},
pages = {1220--1225},
year = {2013},
doi = {},
url = {}
}
DownloadAntonucci, A., Huber, D., Zaffalon, M., Luginbuehl, P., Chapman, I., Ladouceur, R. (2013). CREDO: a military decision-support system based on credal networks. In Proceedings of the 16th Conference on Information Fusion (FUSION 2013), pp. 1–8.
CREDO: a military decision-support system based on credal networks
Authors: Antonucci, A. and Huber, D. and Zaffalon, M. and Luginbuehl, P. and Chapman, I. and Ladouceur, R.
Year: 2013
Abstract: A software tool especially designed for military domains to create and query decision-support systems is presented. Credal networks, which are Bayesian networks whose parameters have the freedom to vary in convex sets, are used to model the relations among the system variables. A novel elicitation procedure of these sets, which allows the military experts to report their knowledge by purely qualitative judgements, is proposed. Two high-level fusion procedures to cope with multiple experts in this framework are also derived. All these features are supported by the software and demonstrated in an application to space security tested during the last NATO multinational experiment.
Published in Proceedings of the 16th Conference on Information Fusion (FUSION 2013), pp. 1–8.
CREDO: a military decision-support system based on credal networks
@INPROCEEDINGS{antonucci2013c,
title = {{CREDO}: a military decision-support system based on credal networks},
booktitle = {Proceedings of the 16th Conference on Information Fusion ({FUSION} 2013)},
author = {Antonucci, A. and Huber, D. and Zaffalon, M. and Luginbuehl, P. and Chapman, I. and Ladouceur, R.},
pages = {1--8},
year = {2013},
doi = {},
url = {}
}
DownloadAntonucci, A., de Rosa, R., Giusti, A., Cuzzolin, F. (2013). Temporal data classification by imprecise dynamical models. In Cozman, F.G., Denoeux, T., Destercke, S., Seidenfeld, T. (Eds), ISIPTA '13: Proceedings of the Eighth International Symposium on Imprecise Probability: Theories and Applications, SIPTA, pp. 13–22.
Temporal data classification by imprecise dynamical models
Authors: Antonucci, A. and de Rosa, R. and Giusti, A. and Cuzzolin, F.
Year: 2013
Abstract: We propose a new methodology to classify temporal data with imprecise hidden Markov models. For each sequence we learn a different model by coupling the EM algorithm with the imprecise Dirichlet model. As a model descriptor, we consider the expected value of the observable variable in the limit of stationar- ity of the Markov chain. In the imprecise case, only the bounds of this descriptor can be evaluated. In practice the sequence, which can be regarded as a trajectory in the feature space, is summarized by a hyperbox in the same space. We classify these static but interval-valued data by a credal generalization of the k-nearest neighbors algorithm. Experiments on benchmark datasets for computer vision show that the method achieves the required robustness whilst outperforming other precise and imprecise methods.
Published in Cozman, F.G., Denoeux, T., Destercke, S., Seidenfeld, T. (Eds), ISIPTA '13: Proceedings of the Eighth International Symposium on Imprecise Probability: Theories and Applications, SIPTA, pp. 13–22.
Temporal data classification by imprecise dynamical models
@INPROCEEDINGS{antonucci2013b,
title = {Temporal data classification by imprecise dynamical models},
editor = {Cozman, F.G. and Denoeux, T. and Destercke, S. and Seidenfeld, T.},
publisher = {SIPTA},
booktitle = {{ISIPTA} '13: Proceedings of the Eighth International Symposium on Imprecise Probability: Theories and Applications},
author = {Antonucci, A. and de Rosa, R. and Giusti, A. and Cuzzolin, F.},
pages = {13--22},
year = {2013},
doi = {},
url = {http://www.sipta.org/isipta13/proceedings/papers/s002.pdf}
}
DownloadAzzimonti, L., Ieva, F., Paganoni, A.M. (2013). Nonlinear nonparametric mixed-effects models for unsupervised classification. Computational Statistics 28(4), pp. 1549–1570.
Nonlinear nonparametric mixed-effects models for unsupervised classification
Authors: Azzimonti, L. and Ieva, F. and Paganoni, A.M.
Year: 2013
Abstract: In this work we propose a novel EM method for the estimation of nonlinear nonparametric mixed-effects models, aimed at unsupervised classification. We perform simulation studies in order to evaluate the algorithm performance and we apply this new procedure to a real dataset.
Published in Computational Statistics 28(4), pp. 1549–1570.
Nonlinear nonparametric mixed-effects models for unsupervised classification
@ARTICLE{azzimonti2013a,
title = {Nonlinear nonparametric mixed-effects models for unsupervised classification},
journal = {Computational Statistics},
volume = {28},
author = {Azzimonti, L. and Ieva, F. and Paganoni, A.M.},
number = {4},
pages = {1549--1570},
year = {2013},
doi = {10.1007/s00180-012-0366-5},
url = {}
}
DownloadBenavoli, A. (2013). The generalised moment-based filter. Automatic Control, IEEE Transactions on 58(10), pp. 2642–2647.
The generalised moment-based filter
Authors: Benavoli, A.
Year: 2013
Abstract: Can we solve the filtering problem from the only knowledge of few moments of the noise terms? In this paper, by exploiting set of distributions based filtering, we solve this problem without introducing additional assumptions on the distributions of the noises (e.g., Gaussianity) or on the final form of the estimator (e.g., linear estimator). Given the moments (e.g., mean and variance) of random variable X, it is possible to define the set of all distributions that are compatible with the moments information. This set can be equivalently characterized by its extreme distributions: a family of mixtures of Dirac deltas. The lower and upper expectation of any function g of X are obtained in correspondence of these extremes and can be computed by solving a linear programming problem. The filtering problem can then be solved by running iteratively this linear programming problem. In this paper, we discuss theoretical properties of this filter, we show the connection with set-membership estimation and its practical applications.
Accepted in Automatic Control, IEEE Transactions on 58(10), pp. 2642–2647.
The generalised moment-based filter
@ARTICLE{benavoli2013b,
title = {The generalised moment-based filter},
journal = {Automatic Control, {IEEE} Transactions on},
volume = {58},
author = {Benavoli, A.},
number = {10},
pages = {2642--2647},
year = {2013},
doi = {10.1109/TAC.2013.2255971},
url = {}
}
DownloadBenavoli, A. (2013). Imprecise hierarchical Dirichlet model with applications. In Information Fusion (fusion), 2013 Proc. Of the 16th International Conference on, pp. 1918–1925.
Imprecise hierarchical Dirichlet model with applications
Authors: Benavoli, A.
Year: 2013
Abstract: Many estimation problems in data fusion involve multiple parameters that can be related in some way by the structure of the problem. This implies that a joint probabilistic model for these parameters should reflect this dependence. In parametric estimation, a Bayesian way to account for this possible dependence is to use hierarchical models, in which data depends on hidden parameters that in turn depend on hyperprior parameters. An issue in this analysis is how to choose the hyperprior in case of lack of prior information. This paper focuses on parametric estimation problems involving multinomial-Dirichlet models and presents a model of prior ignorance for the hyperparameters. This model consists to a set of Dirichlet distributions that expresses a condition of prior ignorance. We analyse the theoretical properties of this model and we apply it to practical fusion problems: (i) the estimate of the packet drop rate in a centralized sensor network; (ii) the estimate of the transition probabilities for a multiple-model algorithm.
Published in Information Fusion (fusion), 2013 Proc. Of the 16th International Conference on, pp. 1918–1925.
Imprecise hierarchical Dirichlet model with applications
@INPROCEEDINGS{benavoli2013d,
title = {Imprecise hierarchical {D}irichlet model with applications},
booktitle = {Information Fusion ({f}usion), 2013 Proc. Of the 16th International Conference on},
author = {Benavoli, A.},
pages = {1918--1925},
year = {2013},
doi = {},
url = {http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=6641239&isnumber=6641065}
}
DownloadBenavoli, A., Papi, F. (2013). Set-membership PHD filter. In Information Fusion (fusion), 2013 Proc. Of the 16th International Conference on, pp. 1722–1729.
Set-membership PHD filter
Authors: Benavoli, A. and Papi, F.
Year: 2013
Abstract: The paper proposes a novel Probability Hypothesis Density (PHD) filter for linear system in which initial state, process and measurement noises are only known to be bounded (they can vary on compact sets, e.g., polytopes). This means that no probabilistic assumption is imposed on the distributions of initial state and noises besides the knowledge of their supports. These are the same assumptions that are used in set-membership estimation. By exploiting a formulation of set-membership estimation in terms of set of probability measures, we derive the equations of the set-membership PHD filter, which consist in propagating in time compact sets that include with guarantee the targets' states. Numerical simulations show the effectiveness of the proposed approach and the comparison with a sequential Monte Carlo PHD filter which instead assumes that initial state and noises have uniform distributions.
Published in Information Fusion (fusion), 2013 Proc. Of the 16th International Conference on, pp. 1722–1729.
Set-membership PHD filter
@INPROCEEDINGS{benavoli2013c,
title = {Set-membership {PHD} filter},
booktitle = {Information Fusion ({f}usion), 2013 Proc. Of the 16th International Conference on},
author = {Benavoli, A. and Papi, F.},
pages = {1722--1729},
year = {2013},
doi = {},
url = {http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=6641211&isnumber=6641065}
}
DownloadBenavoli, A., Zaffalon, M. (2013). Density-ratio robustness in dynamic state estimation. Mechanical Systems and Signal Processing 37(1–2), pp. 54–75.
Density-ratio robustness in dynamic state estimation
Authors: Benavoli, A. and Zaffalon, M.
Year: 2013
Abstract: The filtering problem is addressed by taking into account imprecision in the knowledge about the probabilistic relationships involved. Imprecision is modelled in this paper by a particular closed convex set of probabilities that is known with the name of density ratio class or constant odds-ratio (COR) model. The contributions of this paper are the following. First, we shall define an optimality criterion based on the squared-loss function for the estimates derived from a general closed convex set of distributions. Second, after revising the properties of the density ratio class in the context of parametric estimation, we shall extend these properties to state estimation accounting for system dynamics. Furthermore, for the case in which the nominal density of the COR model is a multivariate Gaussian, we shall derive closed-form solutions for the set of optimal estimates and for the credible region. Third, we discuss how to perform Monte Carlo integrations to compute lower and upper expectations from a COR set of densities. Then we shall derive a procedure that, employing Monte Carlo sampling techniques, allows us to propagate in time both the lower and upper state expectation functionals and, thus, to derive an efficient solution of the filtering problem. Finally, we empirically compare the proposed estimator with the Kalman filter. This shows that our solution is more robust to the presence of modelling errors in the system and that, hence, appears to be a more realistic approach than the Kalman filter in such a case.
Published in Mechanical Systems and Signal Processing 37(1–2), pp. 54–75.
Density-ratio robustness in dynamic state estimation
@ARTICLE{benavoli2012g,
title = {Density-ratio robustness in dynamic state estimation},
journal = {Mechanical Systems and Signal Processing},
volume = {37},
author = {Benavoli, A. and Zaffalon, M.},
number = {1--2},
pages = {54--75},
year = {2013},
doi = {10.1016/j.ymssp.2012.09.004},
url = {http://www.idsia.ch/~alessio/benavoli2012g.pdf}
}
Downloadde Campos, C.P., Cozman, F.G. (2013). Complexity of inferences in polytree-shaped semi-qualitative probabilistic networks. In Proceedings of the 27th AAAI Conference on Advances in Artificial Intelligence (AAAI), pp. 217–223.
Complexity of inferences in polytree-shaped semi-qualitative probabilistic networks
Authors: de Campos, C.P. and Cozman, F.G.
Year: 2013
Abstract: Semi-qualitative probabilistic networks (SQPNs) merge two important graphical model formalisms: Bayesian networks and qualitative probabilistic networks. They provide a very general modeling framework by allowing the combination of numeric and qualitative assessments over a discrete domain, and can be compactly encoded by exploiting the same factorization of joint probability distributions that are behind the Bayesian networks. This paper explores the computational complexity of semi-qualitative probabilistic networks, and takes the polytree-shaped networks as its main target. We show that the inference problem is coNP-Complete for binary polytrees with multiple observed nodes. We also show that inferences can be performed in linear time if there is a single observed node, which is a relevant practical case. Because our proof is constructive, we obtain an efficient linear time algorithm for SQPNs under such assumptions. To the best of our knowledge, this is the first exact polynomial-time algorithm for SQPNs. Together these results provide a clear picture of the inferential complexity in polytree-shaped SQPNs.
Published in Proceedings of the 27th AAAI Conference on Advances in Artificial Intelligence (AAAI), pp. 217–223.
Complexity of inferences in polytree-shaped semi-qualitative probabilistic networks
@INPROCEEDINGS{decampos2013a,
title = {Complexity of inferences in polytree-shaped semi-qualitative probabilistic networks},
booktitle = {Proceedings of the 27th {AAAI} Conference on Advances in Artificial Intelligence ({AAAI})},
author = {de Campos, C.P. and Cozman, F.G.},
pages = {217--223},
year = {2013},
doi = {},
url = {}
}
Downloadde Campos, C.P., Rancoita, P.M.V., Kwee, I., Zucca, E., Zaffalon, M., Bertoni, F. (2013). Discovering subgroups of patients from DNA copy number data using NMF on compacted matrices. PLoS ONE 8(11), e79720.
Discovering subgroups of patients from DNA copy number data using NMF on compacted matrices
Authors: de Campos, C.P. and Rancoita, P.M.V. and Kwee, I. and Zucca, E. and Zaffalon, M. and Bertoni, F.
Year: 2013
Abstract: In the study of complex genetic diseases, the identification of subgroups of patients sharing similar genetic characteristics represents a challenging task, for example, to improve treatment decision. One type of genetic lesion, frequently investigated in such disorders, is the change of the DNA copy number (CN) at specific genomic traits. Non-negative Matrix Factorization (NMF) is a standard technique to reduce the dimensionality of a data set and to cluster data samples, while keeping its most relevant information in meaningful components. Thus, it can be used to discover subgroups of patients from CN profiles. It is however computationally impractical for very high dimensional data, such as CN microarray data. Deciding the most suitable number of subgroups is also a challenging problem. The aim of this work is to derive a procedure to compact high dimensional data, in order to improve NMF applicability without compromising the quality of the clustering. This is particularly important for analyzing high-resolution microarray data. Many commonly used quality measures, as well as our own measures, are employed to decide the number of subgroups and to assess the quality of the results. Our measures are based on the idea of identifying robust subgroups, inspired by biologically/clinically relevance instead of simply aiming at well-separated clusters. We evaluate our procedure using four real independent data sets. In these data sets, our method was able to find accurate subgroups with individual molecular and clinical features and outperformed the standard NMF in terms of accuracy in the factorization fitness function. Hence, it can be useful for the discovery of subgroups of patients with similar CN profiles in the study of heterogeneous diseases.
Published in PLoS ONE 8(11), e79720.
Discovering subgroups of patients from DNA copy number data using NMF on compacted matrices
@ARTICLE{decampos2013d,
title = {Discovering subgroups of patients from {DNA} copy number data using {NMF} on compacted matrices},
journal = {{PLoS} {ONE}},
volume = {8},
author = {de Campos, C.P. and Rancoita, P.M.V. and Kwee, I. and Zucca, E. and Zaffalon, M. and Bertoni, F.},
number = {11},
pages = {e79720},
year = {2013},
doi = {10.1371/journal.pone.0079720},
url = {}
}
DownloadCorani, G., Magli, M., Giusti, A., Gianaroli, L., Gambardella, L. (2013). A Bayesian network model for predicting pregnancy after in vitro fertilization. Computers in Biology and Medicine 43(11), pp. 1783–1792.
A Bayesian network model for predicting pregnancy after in vitro fertilization
Authors: Corani, G. and Magli, M. and Giusti, A. and Gianaroli, L. and Gambardella, L.
Year: 2013
Abstract: We present a Bayesian network model for predicting the outcome of in vitro
fertilization (IVF). The problem is characterized by a particular missingness process; we propose a simple but effective averaging approach which improves parameter estimates compared to the traditional MAP estimation. We present results with generated data and the analysis of a real data set. Moreover, we assess by means of a simulation study the effectiveness of the model in supporting the selection of the embryos to be transferred.
Published in Computers in Biology and Medicine 43(11), pp. 1783–1792.
A Bayesian network model for predicting pregnancy after in vitro fertilization
@ARTICLE{corani2013c,
title = {A {B}ayesian network model for predicting pregnancy after in vitro fertilization},
journal = {Computers in Biology and Medicine},
volume = {43},
author = {Corani, G. and Magli, M. and Giusti, A. and Gianaroli, L. and Gambardella, L.},
number = {11},
pages = {1783--1792},
year = {2013},
doi = {10.1016/j.compbiomed.2013.07.035},
url = {}
}
DownloadCorani, G., Mignatti, A. (2013). Credal model averaging of logistic regression for modeling the distribution of marmot burrows. In Cozman, F.G., Denoeux, T., Destercke, S., Seidenfeld, T. (Eds),, pp. 233–243.
Credal model averaging of logistic regression for modeling the distribution of marmot burrows
Authors: Corani, G. and Mignatti, A.
Year: 2013
Abstract: Bayesian model averaging (BMA) weights the inferences produced by a set of competing models, using as weights the models posterior probabilities. An open problem of BMA is how to set the prior probability of the models. Credal model averaging (CMA) is a credal ensemble of Bayesian models, which generalizes BMA by substituting the single prior over the models by a set of priors. The base models of the ensemble are learned in a Bayesian fashion. We use CMA to ensemble base classi ers which are Bayesian logistic regressors, characterized by dierent sets of covariates. CMA returns indeterminate classi cations when the classi cation is prior-dependent, namely when the most probable class depends on the prior probability assigned to the different models. We apply CMA for modelling the presence and absence of marmot burrows in an Alpine valley in Italy and show that it compares favorably to BMA.
Published in Cozman, F.G., Denoeux, T., Destercke, S., Seidenfeld, T. (Eds), Proceedings of ISIPTA '13 (the Eighth International Symposium on Imprecise Probability: Theories and Applications), pp. 233–243.
Credal model averaging of logistic regression for modeling the distribution of marmot burrows
@INPROCEEDINGS{corani2013a,
title = {Credal model averaging of logistic regression for modeling the distribution of marmot burrows},
journal = {Proceedings of {ISIPTA} '13 (the Eighth International Symposium on Imprecise Probability: Theories and Applications)},
editor = {Cozman, F.G. and Denoeux, T. and Destercke, S. and Seidenfeld, T. },
author = {Corani, G. and Mignatti, A.},
pages = {233--243},
year = {2013},
doi = {},
url = {}
}
DownloadGianaroli, L., Magli, M.C., Gambardella, L., Giusti, A., Grugnetti, C., Corani, G. (2013). Objective way to support embryo transfer: a probabilistic decision. Human Reproduction 28(5), pp. 1210–1220.
Objective way to support embryo transfer: a probabilistic decision
Authors: Gianaroli, L. and Magli, M.C. and Gambardella, L. and Giusti, A. and Grugnetti, C. and Corani, G.
Year: 2013
Abstract: study question: Is it feasible to identify factors that significantly affect the clinical outcome of IVF-ICSI cycles and use them to reliably
design a predictor of implantation?
summary answer: The Bayesian network (BN) identified top-history embryos, female age and the insemination technique as the
most relevant factors for predicting the occurrence of pregnancy (AUC, area under curve, of 0.72). In addition, it could discriminate
between no implantation and single or twin implantations in a prognostic model that can be used prospectively.
what is known already: The key requirement for achieving a single live birth in an IVF-ICSI cycle is the capacity to estimate
embryo viability in relation to maternal receptivity. Nevertheless, the lack of a strong predictor imposes several restrictions on this strategy.
study design, size, duration: Medical histories, laboratory data and clinical outcomes of all fresh transfer cycles performed at the International Institute for Reproductive Medicine of Lugano, Switzerland, in the period 2006–2008 (n 1⁄4 388 cycles), were retrospectively evaluated and analyzed.
participants/materials, setting, methods: Patients were unselected for age, sperm parameters or other infertility cri-
teria. Before being admitted to treatment, uterine anomalies were excluded by diagnostic hysteroscopy.
To evaluate the factors possibly related to embryo viability and maternal receptivity, the class variable was categorized as pregnancy versus
no pregnancy and the features included: female age, number of previous cycles, insemination technique, sperm of proven fertility, the number
of transferred top-history embryos, the number of transferred top-quality embryos, the number of follicles .14 mm and the level of estradiol
on the day of HCG administration. To assess the classifier, the indicators of performance were computed by cross-validation. Two statistical
models were used: the decision tree and the BN.
main results and the role of choice: The decision tree identified the number of transferred top-history embryos, female
age and the insemination technique as the features discriminating between pregnancy and no pregnancy. The model achieved an accuracy of
81.5% that was significantly higher in comparison with the trivial classifier, but the increase was so modest that the model was clinically
useless for predictions of pregnancy. The BN could more reliably predict the occurrence of pregnancy with an AUC of 0.72, and confirmed
the importance of top-history embryos, female age and insemination technique in determining implantation. In addition, it could discriminate
between no implantation, single implantation and twin implantation with the AUC of 0.72, 0.64 and 0.83, respectively.
limitations, reasons for caution: The relatively small sample of the study did not permit the inclusion of more features that
could also have a role in determining the clinical outcome. The design of this study was retrospective to identify the relevant features; a
prospective study is now needed to verify the validity of the model.
wider implications of the findings: The resulting predictive model can discriminate with reasonable reliability between
pregnancy and no pregnancy, and can also predict the occurrence of a single pregnancy or multiple pregnancy. This could represent an ef-
fective support for deciding how many embryos and which embryos to transfer for each couple. Due to its flexibility, the number of variables
in the predictor can easily be increased to include other features that may affect implantation.
study funding/competing interests: This study was supported by a grant, CTI Medtech Project Number: 9707.1 PFLS-L,
Swiss Confederation. No competing interests are declared.
Published in Human Reproduction 28(5), pp. 1210–1220.
Objective way to support embryo transfer: a probabilistic decision
@ARTICLE{corani2013d,
title = {Objective way to support embryo transfer: a probabilistic decision},
journal = {Human Reproduction},
volume = {28},
author = {Gianaroli, L. and Magli, M.C. and Gambardella, L. and Giusti, A. and Grugnetti, C. and Corani, G.},
number = {5},
pages = {1210--1220},
year = {2013},
doi = {10.1093/humrep/det030},
url = {}
}
Downloadvon Hohenstaufen, K.A., Conconi, A., de Campos, C.P., Franceschetti, S., Bertoni, F., Margiotta Casaluci, G., Stathis, A., Ghielmini, M., Stussi, G., Cavalli, F., Gaidano, G., Zucca, E. (2013). Prognostic impact of monocyte count at presentation in mantle cell lymphoma. British Journal of Haematology 162(4), pp. 465–473.
Prognostic impact of monocyte count at presentation in mantle cell lymphoma
Authors: von Hohenstaufen, K.A. and Conconi, A. and de Campos, C.P. and Franceschetti, S. and Bertoni, F. and Margiotta Casaluci, G. and Stathis, A. and Ghielmini, M. and Stussi, G. and Cavalli, F. and Gaidano, G. and Zucca, E.
Year: 2013
Abstract: Increased number of circulating monocytes at presentation has been recently associated with shorter survival in Hodgkin lymphoma, follicular lymphoma and diffuse large B cell lymphoma. This study aimed to assess the prognostic impact of the absolute monocyte count (AMC) at diagnosis in mantle cell lymphoma (MCL). From the series of MCL cases recorded on the databases of the Oncology Institute of Southern Switzerland in Bellinzona (Switzerland) and the Division of Haematology of the Amedeo Avogadro University of Eastern Piedmont in Novara (Italy), the AMC at diagnosis was available in 97 cases. Cox regression was used for both univariate and multivariate analysis. With a median follow up of 7 years, the 5-year overall survival (OS) was 29% for patients with AMC >500/ul and 62% for patients with AMC <= 500/ul (p=0.006). Elevated AMC and beta-2 microglobulin at diagnosis remained independent outcome predictors at multivariate analysis and might be used to build a simple prognostic scoring system. Survival was significantly shorter in patients with both AMC and beta-2 microglobulin above the upper limit of normal but the MCL international prognostic index (MIPI) remained the strongest survival predictor in this series. In this relatively small and heterogeneous series an increased AMC identified poor-risk patients. Our results suggest that AMC in conjunction with the beta-2 microglobulin level might provide an inexpensive way to stratify the MCL patient risk as a complement to the MIPI, which was confirmed to be a very powerful prognostic tool.
Published in British Journal of Haematology 162(4), Blackwell Publishing Ltd, pp. 465–473.
Prognostic impact of monocyte count at presentation in mantle cell lymphoma
@ARTICLE{decampos2013c,
title = {Prognostic impact of monocyte count at presentation in mantle cell lymphoma},
journal = {British Journal of Haematology},
publisher = {Blackwell Publishing Ltd},
volume = {162},
author = {von Hohenstaufen, K.A. and Conconi, A. and de Campos, C.P. and Franceschetti, S. and Bertoni, F. and Margiotta Casaluci, G. and Stathis, A. and Ghielmini, M. and Stussi, G. and Cavalli, F. and Gaidano, G. and Zucca, E.},
number = {4},
pages = {465--473},
year = {2013},
doi = {10.1111/bjh.12409},
url = {http://onlinelibrary.wiley.com/doi/10.1111/bjh.12409/pdf}
}
DownloadMangili, F., Benavoli, A. (2013). New prior near-ignorance models on the simplex. In Cozman, F.G., Denoeux, T., Destercke, S., Seidenfeld, T. (Eds), ISIPTA '13: Proceedings of the Eighth International Symposium on Imprecise Probability: Theories and Applications, SIPTA, Compiegne (FR), pp. 1–9.
New prior near-ignorance models on the simplex
Authors: Mangili, F. and Benavoli, A.
Year: 2013
Abstract: The aim of this paper is to derive new near-ignorance models on the probability simplex, which do not directly involve the Dirichlet distribution and, thus, that are alternative to the Imprecise Dirichlet Model. We focus our investigation to a particular class of distributions on the simplex which is known as the class of Normalized Infinitely Divisible distributions; it includes the Dirichlet distribution as a particular case. Starting from three members of this class, which admit a closed-form expression for the probability density function, we derive three new near-ignorance prior models on the simplex, we analyse their properties and compare them with the Imprecise Dirichlet Model.
Published in Cozman, F.G., Denoeux, T., Destercke, S., Seidenfeld, T. (Eds), ISIPTA '13: Proceedings of the Eighth International Symposium on Imprecise Probability: Theories and Applications, SIPTA, Compiegne (FR), pp. 1–9.
New prior near-ignorance models on the simplex
@INPROCEEDINGS{mangili2013a,
title = {New prior near-ignorance models on the simplex},
editor = {Cozman, F.G. and Denoeux, T. and Destercke, S. and Seidenfeld, T.},
publisher = {SIPTA},
address = {Compiegne (FR)},
booktitle = {{ISIPTA };'13: Proceedings of the Eighth International Symposium on Imprecise Probability: Theories and Applications},
author = {Mangili, F. and Benavoli, A.},
pages = {1--9},
year = {2013},
doi = {},
url = {}
}
DownloadMauá, D.D., de Campos, C.P., Benavoli, A., Antonucci, A. (2013). On the complexity of strong and epistemic credal networks. In Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence, AUAI Press, pp. 391–400.
On the complexity of strong and epistemic credal networks
Authors: Mauá, D.D. and de Campos, C.P. and Benavoli, A. and Antonucci, A.
Year: 2013
Abstract: Credal networks are graph-based statistical models whose parameters take values in a set, instead of being sharply specified as in traditional statistical models (e.g., Bayesian networks). The computational complexity of inferences on such models depends on the irrelevance/independence concept adopted. In this paper, we study inferential complexity under the concepts of epistemic irrelevance and strong independence. We show that inferences under strong independence are NP-hard even in trees with ternary variables. We prove that under epistemic irrelevance the polynomial time complexity of inferences in credal trees is not likely to extend to more general models (e.g. singly connected networks). These results clearly distinguish networks that admit efficient inferences and those where inferences are most likely hard, and settle several open questions regarding computational complexity.
Published in Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence, AUAI Press, pp. 391–400.
On the complexity of strong and epistemic credal networks
@INPROCEEDINGS{maua2013a,
title = {On the complexity of strong and epistemic credal networks},
publisher = {AUAI Press},
booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence},
author = {Mau\'a, D.D. and de Campos, C.P. and Benavoli, A. and Antonucci, A.},
pages = {391--400},
year = {2013},
doi = {},
url = {}
}
DownloadMauá, D.D., de Campos, C.P., Zaffalon, M. (2013). On the complexity of solving polytree-shaped limited memory influence diagrams with binary variables. Artificial Intelligence 205, pp. 30–38.
On the complexity of solving polytree-shaped limited memory influence diagrams with binary variables
Authors: Mauá, D.D. and de Campos, C.P. and Zaffalon, M.
Year: 2013
Abstract: Influence diagrams are intuitive and concise representations of structured decision problems. When the problem is non-Markovian, an optimal strategy can be exponentially large in the size of the diagram. We can avoid the inherent intractability by constraining the size of admissible strategies, giving rise to limited memory influence diagrams. A valuable question is then how small do strategies need to be to enable efficient optimal planning. Arguably, the smallest strategies one can conceive simply prescribe an action for each time step, without considering past decisions or observations. Previous work has shown that finding such optimal strategies even for polytree-shaped diagrams with ternary variables and a single value node is NP-hard, but the case of binary variables was left open. In this paper we address such a case, by first noting that optimal strategies can be obtained in polynomial time for polytree-shaped diagrams with binary variables and a single value node. We then show that the same problem is NP-hard if the diagram has multiple value nodes. These two results close the fixed-parameter complexity analysis of optimal strategy selection in influence diagrams parametrized by the shape of the diagram, the number of value nodes and the maximum variable cardinality.
Published in Artificial Intelligence 205, pp. 30–38.
On the complexity of solving polytree-shaped limited memory influence diagrams with binary variables
@ARTICLE{maua2013b,
title = {On the complexity of solving polytree-shaped limited memory influence diagrams with binary variables},
journal = {Artificial Intelligence},
volume = {205},
author = {Mau\'a, D.D. and de Campos, C.P. and Zaffalon, M.},
pages = {30--38},
year = {2013},
doi = {10.1016/j.artint.2013.10.002},
url = {}
}
DownloadMiranda, E., Zaffalon, M. (2013). Conglomerable coherent lower previsions. In Kruse, R., Berthold, M. R., Moewes, C., Gil, M. A., Grzegorzewski, P., Hryniewicz, O. (Eds), Synergies of Soft Computing and Statistics for Intelligent Data Analysis, Advances in Intelligent and Soft Computing 190, Springer Berlin Heidelberg, pp. 419–427.
Conglomerable coherent lower previsions
Authors: Miranda, E. and Zaffalon, M.
Year: 2013
Abstract: Walley's theory of coherent lower previsions builds upon the former theory by Williams with the explicit aim to make it deal with conglomerability. We show that such a construction has been only partly successful because Walley's founding axiom of joint coherence does not entirely capture the implications of conglomerability. As a way to fully achieve Walley's original aim, we propose then the new theory of conglomerable coherent lower previsions. We show that Walley's theory coincides with ours when all conditioning events have positive lower probability, or when conditioning partitions are nested.
Published in Kruse, R., Berthold, M. R., Moewes, C., Gil, M. A., Grzegorzewski, P., Hryniewicz, O. (Eds), Synergies of Soft Computing and Statistics for Intelligent Data Analysis, Advances in Intelligent and Soft Computing 190, Springer Berlin Heidelberg, pp. 419–427.
Conglomerable coherent lower previsions
@INCOLLECTION{zaffalon2012a,
title = {Conglomerable coherent lower previsions},
editor = {Kruse, R. and Berthold, M. R. and Moewes, C. and Gil, M. A. and Grzegorzewski, P. and Hryniewicz, O.},
publisher = {Springer Berlin Heidelberg},
series = {Advances in Intelligent and Soft Computing},
volume = {190},
booktitle = {Synergies of Soft Computing and Statistics for Intelligent Data Analysis},
author = {Miranda, E. and Zaffalon, M.},
pages = {419--427},
year = {2013},
doi = {10.1007/978-3-642-33042-1_45},
url = {}
}
DownloadMiranda, E., Zaffalon, M. (2013). Conglomerable coherence. International Journal of Approximate Reasoning 54(9), pp. 1322–1350.
Conglomerable coherence
Authors: Miranda, E. and Zaffalon, M.
Year: 2013
Abstract: We contrast Williams' and Walley's theories of coherent lower previsions in the light of conglomerability. These are two of the most credited approaches to a behavioural theory of imprecise probability. Conglomerability is the notion that distinguishes them the most: Williams' theory does not consider it, while Walley aims at embedding it in his theory. This question is important, as conglomerability is a major point of disagreement at the foundations of probability, since it was first defined by de Finetti in 1930. We show that Walley's notion of joint coherence (which is the single axiom of his theory) for conditional lower previsions does not take all the implications of conglomerability into account. Considered also some previous results in the literature, we deduce that Williams' theory should be the one to use when conglomerability is not required; for the opposite case, we define the new theory of conglomerably coherent lower previsions, which is arguably the one to use, and of which Walley's theory can be understood as an approximation. We show that this approximation is exact in two important cases: when all conditioning events have positive lower probability, and when conditioning partitions are nested.
Published in International Journal of Approximate Reasoning 54(9), pp. 1322–1350.
Conglomerable coherence
@ARTICLE{zaffalon2013b,
title = {Conglomerable coherence},
journal = {International Journal of Approximate Reasoning},
volume = {54},
author = {Miranda, E. and Zaffalon, M.},
number = {9},
pages = {1322--1350},
year = {2013},
doi = {10.1016/j.ijar.2013.04.016},
url = {}
}
DownloadMiranda, E., Zaffalon, M. (2013). Computing the conglomerable natural extension. In Cozman, F., Denoeux, T., Destercke, S., Seidenfeld, T. (Eds), ISIPTA '13: Proceedings of the Eighth International Symposium on Imprecise Probability: Theories and Applications, SIPTA, pp. 255–264.
Computing the conglomerable natural extension
Authors: Miranda, E. and Zaffalon, M.
Year: 2013
Abstract: Given a coherent lower prevision P, we consider the problem of computing the smallest coherent lower prevision F ≥ P that is conglomerable, in case it exists. F is called the conglomerable natural extension. Past work has showed that F can be approximated by an increasing sequence (En)n∈ℕ of coherent lower previsions. We close an open problem by showing that this sequence can be made of infinitely many distinct elements. Moreover, we give sufficient conditions, of quite broad applicability, to make sure that the point-wise limit of the sequence is F in case P is the lower envelope of finitely many linear previsions. In addition, we study the question of the existence of F and its relationship with the notion of marginal extension.
Published in Cozman, F., Denoeux, T., Destercke, S., Seidenfeld, T. (Eds), ISIPTA '13: Proceedings of the Eighth International Symposium on Imprecise Probability: Theories and Applications, SIPTA, pp. 255–264.
Computing the conglomerable natural extension
@INPROCEEDINGS{zaffalon2013c,
title = {Computing the conglomerable natural extension},
editor = {Cozman, F. and Denoeux, T. and Destercke, S. and Seidenfeld, T.},
publisher = {SIPTA},
booktitle = {{ISIPTA };'13: Proceedings of the Eighth International Symposium on Imprecise Probability: Theories and Applications},
author = {Miranda, E. and Zaffalon, M.},
pages = {255--264},
year = {2013},
doi = {},
url = {http://www.sipta.org/isipta13/proceedings/papers/s025.pdf}
}
DownloadZaffalon, M., Miranda, E. (2013). Probability and time. Artificial Intelligence 198, pp. 1–51.
Probability and time
Authors: Zaffalon, M. and Miranda, E.
Year: 2013
Abstract: Probabilistic reasoning is often attributed a temporal meaning, in which conditioning is regarded as a normative rule to compute future beliefs out of current beliefs and observations. However, the well-established ‘updating interpretation’ of conditioning is not concerned with beliefs that evolve in time, and in particular with future beliefs. On the other hand, a temporal justification of conditioning was proposed already by De Moivre and Bayes, by requiring that current and future beliefs be consistent. We reconsider the latter approach while dealing with a generalised version of the problem, using a behavioural theory of imprecise probability in the form of coherent lower previsions as well as of coherent sets of desirable gambles, and letting the possibility space be finite or infinite. We obtain that using conditioning is normative, in the imprecise case, only if one establishes future behavioural commitments at the same time of current beliefs. In this case it is also normative that present beliefs be conglomerable, which is a result that touches on a long-term controversy at the foundations of probability. In the remaining case, where one commits to some future behaviour after establishing present beliefs, we characterise the several possibilities to define consistent future assessments; this shows in particular that temporal consistency does not preclude changes of mind. And yet, our analysis does not support that rationality requires consistency in general, even though pursuing consistency makes sense and is useful, at least as a way to guide and evaluate the assessment process. These considerations narrow down in the special case of precise probability, because this formalism cannot distinguish the two different situations illustrated above: it turns out that the only consistent rule is conditioning and moreover that it is not rational to be willing to stick to precise probability while using a rule different from conditioning to compute future beliefs; rationality requires in addition the disintegrability of the present-time probability.
Published in Artificial Intelligence 198, pp. 1–51.
Probability and time
@ARTICLE{zaffalon2013a,
title = {Probability and time},
journal = {Artificial Intelligence},
volume = {198},
author = {Zaffalon, M. and Miranda, E.},
pages = {1--51},
year = {2013},
doi = {10.1016/j.artint.2013.02.005},
url = {}
}
Download 
