Reformulating Inference Problems Through Selective Conditioning

We describe how we selectively reformulate portions of a belief network that pose difficul­ ties for solution with a stochastic-simulation algorithm. With employ the selective con· ditioning approach to target specific nodes in a belief network for decomposition, based on the contribution the nodes make to the tractability of stochastic simulation. We re­ view previous work on BNRAS algorithms­ randomized approximation algorithms for probabilistic inference. We show how selec­ tive conditioning can be employed to refor­ mulate a single BNRAS problem into multiple tractable BNRAS simulation problems. We discuss how we can use another simulation algorithm-logic sampling-to solve a com­ ponent of the inference problem that provides a means for knitting the solutions of individ­ ual subproblems into a final result. Finally, we analyze tradeoffs among the computa­ tional subtasks associated with the selective­ conditioning approach to reformulation.