Bayesian Knowledge Bases
From Distributed Information and Intelligence Analysis Group
A Bayesian knowledge-base is a knowledge representation system that has a similar structure as a Bayesian network. It represents knowledge by an “if-then” structure and represents uncertainty by probability theory. BKBs subsume BNs and are advantageous to BNs because BKBs are robust to knowledge incompleteness and cyclic information, while still preserving a probabilistically sound representation of uncertainty. We now extend BKBs to represent both time and interaction since time is the key component of a dynamic world. We target to provide a comprehensive model that can describe the interactions of knowledge in the world, as well as reason through them.
