Belief formation from observation and belief integration usingvirtual belief space in Dempster-Shafer probability model

摘要

Integrating uncertain information from multiple sources is a keyntechnology to realise reliable AI systems. The Dempster-Shafernprobability model (DS model) provides a useful computational scheme fornthe integration. In this paper, the author proposes two algorithms fornbelief formation and integration based on the DS model. The firstnalgorithm is for computing a basic probability assignment function basednon similarity measures between observed data and object categories. Thensoundness of the algorithm is shown using mathematical relations betweennseveral fuzzy measures. Then, the author proposes a new algorithm fornintegrating multiple beliefs (i.e, basic probability assignmentnfunctions). Using this algorithm, the author can solve a controversialnproblem in the DS model about how to combine partially conflictingnbeliefs. That is, with the proposed algorithm, the author can smoothlynintegrate multiple beliefs even if they are partially/totallynconflicting. From a computational viewpoint, moreover, the beliefnintegration by the proposed algorithm can be implemented verynefficiently