Bounded Conditioning: Flexible Inference for Decisions under Scarce Resources

摘要: We introduce a graceful approach to probabilistic inference called bounded conditioning. Bounded conditioning monotonically refines the bounds on posterior probabilities in belief network with computation, and converges final of interest allocation complete resource fraction. The allows reasoner exchange arbitrary quantities computational for incremental gains quality. As such, holds promise as useful technique reasoning under general conditions uncertain varying resources. algorithm solves bounding problem complex networks by breaking into set mutually exclusive, tractable subproblems ordering their solution expected effect that each subproblem will have answer. algorithm, discuss its characterization, present performance several networks, including model about problems intensive-care medicine.