We are a group of researchers at IDSIA
that share a common interest in imprecise probability
. We are interested in particular in modeling uncertainty and randomness with sets of probabilities (such as, e.g., closed convex sets of probability mass functions or distributions, or probability intervals), but we work also on the more traditional side of precise probability.
We apply these ideas to areas like classification (data mining), filtering and tracking (engineering), knowledge-based systems (artificial intelligence), learning from data (statistics and machine learning), and to real-world problems like defense and surveillance, medical diagnosis, bioinformatics, environmental problems, collaborative filtering.
We work also on the foundations of imprecise probability: on the theories of coherent lower previsions and of coherent sets of desirable gambles; on the problem of prior ignorance when learning from data; and on the related problem of handling incomplete and missing information.
We are one of the main groups responsible for the creation and development of credal networks and credal classification, and more generally speaking we are tightly related to the research on probabilistic graphical models such as Bayesian networks.