ABSTRACT
Probabilistic graphical models provide a useful framework for building parsimonious models for high-dimensional data. They are based on an interplay between probability theory and graph theory, in which the properties of an underlying graph specify the conditional independence properties of a set of random variables. In typical applications, the structure of this graph is not known, and it is of interest to estimate it based on samples, a problem known as graphical model selection. In this chapter, we discuss a variety of methods based on `1-regularization designed for this purpose.
