Relevance-Based Sequential Evidence Processing in Bayesian Networks

1-~.levance reasoning in Bayesian networks can be la~ed to improve efficiency of belief updating ",dgorithaL~ by identit~ing and pruning those parts of a network that are irrelevant ibr the computation. Relevance reaqoning is based on the graphical property of d-separation and other simple and efficient techniques, the computational complexity of which is usuaUy negligible when compared to the complt~xity of belief updating in general. This paper describes a belief updating technique based on relevance reasoning that is applicable in practiced systems ia which observations are interleaved with belief updating. Our technique invalidates the posterior beliefs of those nodes that (h~pentl probabilisdcalb" oa the new evidence ~md focuses the subsequent belief updating on the inv~didated beliefs rather than on all bcliet~. Vexy often observations inwflidate only a small fraction of the beliefs and our scheme can then lead to substaatial savings in computation. We report rcsuks of empirical tests that demomstratc practical signiiicanre of our approach.