One Kind Of Parallel Method Of Learning Bayesian Network

Nowadays, learning Bayesian network is one of the hottest topics inthe domain of Artificial Intelligence. However it is NP-Problem inregards to the structures of the network increasing exponentially with thegrowth of the number of variables and their shifting states. In order toovercome the shortcoming of the sluggish computation and search for therefined networks, a range of researches have been done, and the relatedalgorithms were presented. Unfortunately, they all have their ownlimitations. In response to such sort of problems, meta-heuristic methodstands out and presents a fairly satisfying performance. The ACOalgorithm, which is one of meta-heuristic algorithm, shows excellentefficiency in solving such combination optimization problems. But thelimitation does exist in view of its high complexity of time and space. Itis still long distance to go before the methods are sought out to improvethe performance about its results. With the development of highperformance computing platform, some algorithms of combinationoptimization problems by paralyzed approach come into being gradually.In this paper, a parallel algorithm of PACOB is put forward to solve theproblem of learning Bayesian networks. It sheds new lights on thelearning of Bayesian network by applying the algorithm of BDE approachand MDL approach synergistically.