Towards a Faster Inference Algorithm in Multiply Sectioned Bayesian Networks
Multiply sectioned Bayesian network(MSBN) is an extension of Bayesian network(BN) model for the support of flexible modelling in large and complex problem domains. However, current MSBN inference methods involve extensive intra-subnet(internal) and inter-subnet (external) message passings. In this paper, we present a new MSBN message passing scheme which substantially reduces the total number of message passings. By saving on both internal and external messages, our method improves the overall efficiency of MSBN inference compared with existing methods.
KeywordsBayesian Network Local Propagation Compilation Time Incoming Message Local Consistency
Unable to display preview. Download preview PDF.
- 3.Lepar, V., Shenoy, P.P.: A comparison of Lauritzen-Spiegelhalter, Hugin, and Shenoy-Shafer architectures for computing marginals of probability distributions. In: Proceedings of the 14th Conference on Uncertainty in Artificial Intelligence (UAI 1998), pp. 328–337. Morgan Kaufmann, San Francisco (1998)Google Scholar
- 4.Pearl, J.: Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference. Morgan Kaufmann Publishers, San Francisco, California (1988)Google Scholar
- 5.Shafer, G.: An axiomatic study of computation in hypertrees. School of Business Working Papers 232, University of Kansas (1991)Google Scholar
- 6.Wu, D., Jin, K.: Demystify the messages in the hugin architecture for probabilistic inference and its application. In: FLAIRS Conference, pp. 55–61 (2006)Google Scholar
- 9.Xiang, Y.: Probabilistic Reasoning in Multuagent Systems: A Graphical Models Approach, Cambridge (2002)Google Scholar
- 10.Xiang, Y.: Comparison of multiagent inference methods in multiply sectioned bayesian networks (2003)Google Scholar
- 11.Xiang, Y., Jensen, F.V., Chen, X.: Inference in multiply sectioned bayesian networks: Methods and performance comparison. IEEE Transaction on Systems, Man, and Cybernetics 36, 546–558 (2006)Google Scholar