Exploiting Global Structures in Bayesian Network Compilation by Zero-suppressed BDDs

Abstract. This paper addresses an algorithm for probabilistic inference for Bayesian Networks (BNs). Our algorithm uses Zero-suppressed BDDs for compiling BNs. We introduce “separation variable ” to reflect global structures of BNs, which provides more compact ZBDDs. We show some experimental results to compare our method with the state-of-the-art tool