ABSTRACT
This chapter describes the graphical modelling paradigm which provides simple ways of expressing important properties of the model. A graphical model is a probability or belief function model whose factorization can be represented with a graph. The chapter follows the approach of Kong, assuming the existence of the graphical model and deriving its properties. It contains a brief review of graph theory, and defines the most commonly used graphical models and the relationship between them. The chapter describes certain Markovian conditional independence statements which can be derived from the graphical model and other similar graphical modelling techniques. Most questions asked of graphical models involve either finding margins of the graphical belief function or finding margins of the graphical belief function conditioned on certain observations or hypotheses. A big advantage of using graphical models is that these margins can often be computed by local computations, at a considerable savings of time and space.
