Factor tree inference algorithm (FTI) is an exact inference algorithm for diagnostic Bayesian networks (DBNs). Through computation sharing, the efficiency of FTI can be superior to conventional exact inference algorithms when answering multiple queries. However, there are circumstances where the “factor tree” is cyclic; FTI can not perform in these situations. In this article, we propose a factor graph inference algorithm (FGI). FGI can perform when the factor graph is cyclic, and reduces to FTI when it is acyclic. We demonstrate the benefit of FGI on a real-world DBN.
展开▼
