Localized Partial Evaluation of Belief Networks

Most algorithms for propagating evidence through belief networks have been exact and exhaustive: they produce an exact (point­ valued) marginal probability for every node in the· network. Often, however, an appli­ cation will not need information about ev­ ery node in the network nor will it need ex­ act probabilities. We present the localized partial evaluation (LPE) propagation algo­ rithm, which computes interval bounds on the marginal probability of a specifted query node by examining a subset of the nodes in the entire network. Conceptually, LPE ig­ nores parts of the network that are "too far away" from the queried node to have much impact on its value. LPE has the "anytime" property of being able to produce better so­ lutions (tighter intervals) given more time to consider more of the network.