COMPUTATIONAL PROPERTIES OF 2 EXACT ALGORITHMS FOR BAYESIAN NETWORKS

This paper studies computational properties of two exact inference alg orithms for Bayesian networks, namely the clique tree propagation algo rithm (CTP)(1) and the variable elimination algorithm (VE). VE permits pruning of nodes irrelevant to a query while CTP facilitates sharing of computations among different queries. Experiments have been conduct ed to empirically compare VE and CTP. We found that, contrary to commo n beliefs, VE is often more efficient than CTP, especially in complex networks.