Research Of Parallel VE Inference Algorithm In Bayesian Network

Bayesian network is uncertain knowledge representation model based on the probability theory and graph theory. On the one hand, it directly reveals the structure of the problems with the form of graph theory; on the other hand, it uses the probability principle to deal with the structure of the problem to reduce the computational complexity of inference.In the worst case, the time complexity of inference is exponential computation on the Bayesian network, and the complexity of inference depends on the network structure. So it is the hot spot to design effective inference methods. But the practical system of real world is robustness and requires high performance. The time complexity of existing exact inference algorithm is still high, it has certain limitations to use in the application.The main purpose of parallel computing is to quickly solve large and complex computational problems. This article introduces the parallel computing into the variable elimination algorithm, it can make the variable elimination algorithm be divided into parts to be processed in parallel. It can greatly reduce the time complexity of the variable elimination algorithm and have certain value.The article discusses this topic from three perspectives as following:First, the article researches the basic principles and methods of the Bayesian network and parallel computing, it detailed analysis exact inference algorithms at this stage and their pros and cons. The article focuses on to analysis the variable elimination algorithm.Second, we combine the variable elimination algorithm with the parallel computing technology and consider how to make the variable elimination algorithm process in parallel and how to achieve the algorithm.Finally, we do the detailed experiments and analysis on the parallel variable elimination algorithm. We experiment on two standard Bayesian network to verify efficiency of this algorithm. In conclusion, the article introduces parallel computing into the Bayesian network inference; theoretical analysis and experimental verification proved the parallel variable elimination algorithm is correct and efficient.