INDEPENDENCE OF CAUSAL INFLUENCE AND CLIQUE TREE PROPAGATION

This paper explores the role of independence of causal influence (ICI) in Bayesian network inference. ICI allows one to factorize a conditio nal probability table into smaller pieces. We describe a method for ex ploiting the factorization in clique tree propagation (CTP) - the stat e-of-the-art exact inference algorithm for Bayesian networks (BNS). We also present empirical results showing that the resulting algorithm i s significantly more efficient than the combination of CTP and previou s techniques for exploiting ICI. (C) 1998 Published by Elsevier Scienc e Inc. All rights reserved.