On the Generation of Alternative Explanations with Implications for Belief Revision

In general, the best explanation for a given observation makes no promises on how good it is with respect to other alternative explana­ tions. A major deficiency of message-passing schemes for belief revision in Bayesian net­ works is their inability to generate alterna­ tives beyond the second best. In this pa­ per, we present a general approach based on linear constraint systems that naturally gen­ erates alternative explanations in an orderly and highly efficient manner. This approach is then applied to cost-based abduction prob­ lems as well as belief revision in Bayesian net­ works.