Efficient Subjective Bayesian network belief propagation for trees

Subjective Bayesian networks extend Bayesian networks by incorporating uncertainty in the conditional probabilities. This paper develops subjective belief propagation (SBP) that extends regular belief propagation (BP) to efficiently infer uncertain marginal probabilities in subjective Bayesian networks. It is shown that SBP's runtime exhibits only slightly slower performance than standard BP but is able to effectively characterize a distribution for the marginals. Simulations affirm that unlike the valuation-based system, a previous uncertain probabilistic reasoning framework, SBP is able to effectively capture bounds for the actual error in a consistent manner.