Dynamic Bayesian Networks: Introduction

From DSL

A BN is useful for problem domains where the state of the world is static. In such a world, every variable has a single and fixed value. Unfortunately, this assumption of a static world does not always hold, as many domains exist where variables are dynamic and reasoning over time is necessary, such as dynamic systems.

A Dynamic Bayesian network (DBN) is a BN extended with a temporal dimension to enable us to model dynamic systems [DK88]. This tutorial assumes a basic knowledge about the DBN formalism and GeNIe. If this is not the case, please refer to Bayesian networks and GeNIe tutorials. The temporal extension of BN does not mean that the network structure or parameters changes dynamically, but that a dynamic system is modeled. A DBN is a directed, a-cyclic graphical model of a stochastic process. It consists of time-slices (or time-steps), with each time-slice containing its own variables.

The implementation of DBN in Genie is an extension of Murphy's DBN [Mur02]. To learn more about this extension please refer to[Hulst06].

The following sections give a brief description of a scenario that may require a DBN and how it can be modelled in GeNIe.