Chapter 2. Exact Inference

After building a graphical model, one of the main tasks one wants to perform is putting questions and queries to the model. There are many ways to use graphical models and the representation they give from a joint probability distribution. For example, we can study interactions between random variables. We can also see if any correlation or causality is captured by the model. Moreover, as probability models governing the random variables are parameterized, it means that their probability distribution is fully known through being familiar with some numerical parameters. We might be interested in knowing the values of those parameters when other variables are observed.

The main focus of this chapter is on introducing algorithms to query a distribution that uses the model and observations on a subset of variables in order to discover the posterior probability distribution on another subset. It is not necessary to observe and query all the variables. In fact...