Exponential Family Graphical Models: Inference, Learning and Convexity

Jason K. Johnson and O. Patrick Kreidl

The objective of this presentation is to provide an overview of exponential family graphical models and to introduce the following topics: (i) exact recursive inference methods, (ii) model identification methods and (iii) the role of convex duality in these problems. We also touch upon some advanced topics, including approximate inference and model structure estimation.