[Wiley Series in Probability and Statistics] Bayesian Networks (An Introduction) || Graphical Models and Exponential Families

✍ Scribed by Koski, Timo; Noble, John M.

• Publisher: Wiley
• Year: 2009
• Weight: 628 KB
• Category: Article
• ISBN: 047068402X
• DOI: 10.1002/9780470684023.ch8

✦ Synopsis

This chapter introduces exponential families of distributions, focusing on links with convex analysis and specifically with the theory of conjugate duality. This is then applied to updating the probability distribution in a graphical model in the light of new information. One of the key features of exponential families, with mean parameters, is the relative ease with which the entropy and Kullback-Leibler distance between two members of the family may be computed.