On Tree Structures Used By Simple Propagation

Simple Propagation (SP) is a new junction tree-based algorithm for probabilistic inference in discrete Bayesian networks. It is similar to Lazy Propagation, but uses a simpler approach to exploit the factorization during message computation. The message construction is based on a one-in, one-out-principle meaning a potential has at least one non-evidence variable in the separator and at least one non-evidence variable not in the separator. This paper considers the use of different tree structures to guide the message passing in SP and reports on an experimental analysis using a set of real-world Bayesian networks.