News
07/10/2011
End Date:
13/10/2011
Lecture Announcement
As part of a cycle of short lectures on Stochastic Models, we announce the lecture by Luigi Portinale "PGM: Probabilsitic Graphical Models", Wed. October 12, Seminar Room
Abstract: PGMs are probabilistic models very popular in AI with a wide range of applications: diagnosis, monitoring and recovery, planning, data mining and knowledge discovery, user profiling, recommender systems, etc... They are based on the factorization of the sample space in a set of relevant variables, for which only local dependencies are directly represented. The lecture will concentrate on so called "directed models", discussing some of the most popular formalisms like Bayesian Networks, Influence Diagrams and their dynamic counterparts (Dynamic Bayesian Netowrks and Dynamic Decision Networks). Basic features concerning representations, inference and application tasks will be discussed.