This paper is devoted to the clustering belief updating algorithm using the junction tree as a tree graph representation of Bayesian networks. The algorithm is applicable for predictions based on a learned Bayesian network as well as for supporting an exact network learning process, for example, the EM algorithm. The constructing steps and the principles of work with the junction tree are specified. The software implementation of the algorithm is also considered.

Keywords: Bayesian network, belief network, belief update, expert system, join tree, junction tree, probabilistic interference, probabilistic propagation

• PostScript (in Russian) (6.99MB)
• PDF (in Russian) (0.98MB)



• PostScript. zip (in Russian) (1.80MB)
• PDF. zip (in Russian) (931KB)