Graphical Models

13 Dezembro 2016, 13:00 • João Marques Silva

This week's lecture studied graphical models (also known Bayesian networks). The first part of the lecture covered the syntax and semantics of Bayesian networks, and analyzed several examples were studied. A number of additional topics were studied, that include compact representation of conditional probability tables, and hybrid networks. The second part of the lecture covered inference in Bayesian networks. Exact inference was studied in detailed, including approaches based on model enumeration and variable elimination. Methods for approximate methods were briefly overviewed. The second part of the lecture also covered the proof that inference in Bayesian networks in NP-hard, by a reduction from the 3SAT problem.