Testing and Improving Local Adaptive Importance Sampling in LFJ Local-JT in Multiply Sectioned Bayesian Networks

摘要

Multiply Sectioned Bayesian Network (MSBN) provides a model for probabilistic reasoning inmulti-agent systems. The exact inference is costly and difficult to be applied in the context ofMSBNs as the size of problem domain becomes larger and complex. So the approximatetechniques are used as an alternative in such cases. Recently, for reasoning in MSBNs, LJFbasedLocal Adaptive Importance Sampler (LLAIS) has been developed for approximatereasoning in MSBNs. However, the prototype of LLAIS is tested only on Alarm Network (37nodes). But further testing on larger networks has not been reported yet, so the scalability andreliability of algorithm remains questionable. Hence, we tested LLAIS on three large networks(treated as local JTs) namely Hailfinder (56 nodes), Win95pts (76 nodes) and PathFinder(109nodes). From the experiments done, it is seen that LLAIS without parameters tuned shows goodconvergence for Hailfinder and Win95pts but not for Pathfinder network. Further when theseparameters are tuned the algorithm shows considerable improvement in its accuracy andconvergence for all the three networks tested.