Having taken Professor Darwiche's course on Bayesian Networks, I was excited to get my hands on this book, which is a culmination of the notes from that class and his research on the subject.
This is an excellent text, with very clear explanations and step by step descriptions in pseudo code of the important algorithms in the text.
The first few chapters lay the probabilistic foundations needed for understanding Bayesian Networks and the conditional independences such networks encode.
Chapter 5 gives examples in several different domains of using Bayesian Networks to model different systems and answer queries about them.
After this, the book gets into the meat of its primary focus, efficient probabilistic inference in the context of Bayesian Networks.
It lays out various algorithms for exact inference using jointrees or recursive conditioning, and the complexity and trade-offs of the different approaches.
It further details further refinements that can reduce networks in some cases for even better performance.
After this, it details approximate inference techniques including sampling and belief propagation.
Chapter 14 on belief propagation is especially good, with its discussions on the semantics of belief propagation, generalized belief propagation, and an alternative formulation of generalized belief propagation edge deletion belief propagation.
The last few chapters also delve into learning Bayesian Networks structure and parameters.
All in all, this book will give an in depth knowledge of exact and approximate inference in Bayesian networks and a good overview of learning and applying these models to various domains.