TRANSFORMATION OF SINGLY-CONNECTED BAYESIAN BELIEF NETWORKS INTO ALGEBRAIC BAYESIAN NETWORKS
Saint Petersburg State University, Saint Petersburg, 199034, Russian Federation; St. Petersburg Federal Research Center of the Russian Academy of Sciences, Saint Petersburg, 199178, Russian Federation; Professor; Head of Laboratory
Read the full article
Abstract. The paper presents the key steps of the algorithm which can transform uncertain knowledge patterns bases represented as singly-connected Bayesian belief networks (i.e. structured as a poly-tree) into the bases represented as algebraic Bayesian networks. The major part of this transformation successively performs calculation of join probability tensors based on already-known conditional probability tensors that are stored in the nodes of Bayesian belief networks. The transformation process is completed when join probability tensors are re-calculated into the set of probability estimates over elements of conjuncts ideals.
Keywords:
Bayesian networks, probabilistic semantics, uncertain knowledge models, probabilistic-logic inference.
