On a Distributed Anytime Architecture for Probabilistic Reasoning.
Eugene Santoe, Solomon Eyal Shimony, Williams Solomon E., Edward Edward · 1995
Abstract : An architecture for unifying various algorithms for probabilistic reasoning is presented. Any algorithms having anytime, anywhere characteristics may be mixed in this scheme. Since algorithms for probabilistic reasoning have widely different behavior over classes of Bayes networks, the scheme permits taking advantage of the set of algorithms that happen to perform well for the problem instance at hand. We concentrate on belief updating and belief revision. Some results are presented for our system (OVERMIND) consisting of several genetic algorithm instances, A*, etc. running in parallel.