Belief network algorithms: A study of performance based on domain characterisation

Abstract

In recent years belief networks have become a popular representation for reasoning with incomplete and changing information and are used in a wide variety of applications. There are a number of exact and approximate inference algorithms available for performing belief updating, however in general the task is NP-hard. Typically comparisons are made of only a few algorithms, and on a particular example network. We survey belief network algorithms and propose a system for domain characterisation as a basis for algorithm comparison. We present performance results using this framework from three sets of experiments: (1) on the Likelihood Weighting (LW) and Logic Sampling (LS) stochastic simulation algorithms? (2) on the performance of LW and Jensen's algorithms on state-space abstracted networks, (3) some comparisons of the time performance of LW, LS and the Jensen algorithm. Our results indicate that domain characterisation can be useful for predicting inference algorithm performance on a belief network for a new application domain.