This will be a short course that will serve as a general introduction to graphical models and Bayesian networks. Topics will include conditional independence, Markov equivalence, directed and undirected models, general model semantics, decomposable and triangulated models, graph algorithms, elimination and triangulation methods, triangulated graphs and junction-trees, the message passing protocol, exact probabilistic inference, and issues that arise with dynamic Bayesian networks. This material will be covered at a somewhat thorough level mathematically, going over the theorems and proofs needed to understand the topic well.

• Lauritzen, "Graphical Models"
• Upcoming text by Mike Jordan
• Course notes from the course Jeff Bilmes teaches at UW on this topic.