Fast Bayes and the dynamic junction forest

It has been shown that junction tree algorithms can provide a quick and eff icient method for propagating probabilities in complex multivariate problem s when they can be described by a fixed conditional independence structure. In this paper we formalise and illustrate with two practical examples how these probabilistic propagation algorithms can be applied to high dimension al processes whose conditional independence structure, as well as their und erlying distributions, are augmented through the passage of time. (C) 1999 Elsevier Science B.V. All rights reserved.