Graphical models (Cowell et al., 1999; Bishop, 2006; Neapolitan, 2004; Pearl, 1988; Jensen et al., 1990; Murphy, 2001) provide a formalism for defining the structure for a probabilistic model. A graphical model is a graph whose nodes represent random variables and edges represent direct dependencies. The models presented here vary mostly in whether they are static or dynamic and whether the variables are discrete or continuous valued.
Graphical models have evolved from being a mere academic curiosity into a popular field of research with a huge number of applications. The applications range from network engineering to bioinformatics.
All the models and methods studied in this thesis can be seen as extensions of these basic models, which are therefore introduced here.