Discrete graphical models - An optimization perspective

Abstract

This monograph is about combinatorial optimization. More precisely, about a special class of combinatorial problems known as energy minimization or maximum a posteriori (MAP) inference in graphical models, closely related to weighted and valued constraint satisfaction problems and having tight connections to Markov random fields and quadratic pseudo-boolean optimization. What distinguishes this monograph from a number of other monographs on graphical models is its focus: It considers graphical models, or, more precisely, MAP-inference for graphical models, purely as a combinatorial optimization problem. Modeling, applications, probabilistic interpretations and many other aspects are either ignored here or find their place in examples and remarks only.