A new approach for convert multiply-connected trees in Bayesian networks

Abstract

One of the purposes of the Bayesian networks is inference. There have different algorithms for this purpose. Message passing algorithms us one of the inference approaches on the junction trees. However all of the Bayesian networks are not singly-connected tree, but solution was presented for making junction tree from multiply-connected tree. This article presents a new approach for converting multiply-connected tree to junction tree by moor and meely machine concepts. This approach is based on edge labeling by conditional probability, then moral and triangulation steps transact based on five-step design.