UMBC ebiquity
29th International FLAIRS Conference
Inconsistent Knowledge Integration with Bayesian Network
May 16, 2016
Given a Bayesian network (BN) representing a probabilistic knowledge base of a domain, and a set of low-dimensional probability distributions (also called constraints) representing pieces of new knowledge coming from more up-to-date or more specific observations for a certain perspective of the domain, we present a theoretical framework and related methods for integrating the constraints into the BN, even when these constraints are inconsistent with the structure of the BN due to dependencies among relevant variables in the constraint being absent in the BN.
AAAI Press
Downloads: 388 downloads