Belief update in CLG Bayesian networks with lazy propagation

摘要: In recent years, Bayesian networks with a mixture of continuous and discrete variables have received an increasing level attention. this paper, we focus on the restricted class known as conditional linear Gaussian (CLG networks) present architecture for exact belief update networks. The proposed is extension lazy propagation using operations Lauritzen Jensen [S.L. Lauritzen, F. Jensen, Stable local computation mixed distributions, Statistics Computing 11(2) (2001) 191-203] Cowell [R.G. Cowell, Local in networks, Journal Machine Learning Research 6 (2005) 1517-1550]. By decomposing clique separator potentials into sets factors, takes advantage independence irrelevance properties induced by structure graph evidence. resulting benefits are illustrated examples assessed experiments. performance has been evaluated set randomly generated results indicate significant potential architecture.