next up previous
Next: Bayesian Knowledge Bases Up: Inferencing Over Incomplete Solution Previous: Inferencing Over Incomplete Solution

Introduction

A Bayesian Knowledge Base (BKB) is a knowledge representation for information containing uncertainty. BKBs use random variables(RVs) and probabilities to represent knowledge. Unlike other approaches such as Bayesian Networks[4], BKBs are designed to handle incompleteness in our knowledgegif. One of the inferencing tasks performed on BKBs is belief revision. In belief revision, we attempt to find the most likely state for every variable. Unfortunately, when we use genetic algorithms(GAs)[2, 5, 3] to perform belief revision, there are many combinations of states that occur which are not computable. These combinations occur when the relation between states of RVs is incomplete. If we don't evaluate the incompletenesses, the GA has a relatively flat solution space and performs poorly. In this paper, we provide a solution to this problem by developing a way to evaluate the incompletenesses. By evaluating incompletenesses, we create new slopes for the GA to climb. This enables the GA to converge to the optimal solution faster.

Section 2 provides a description and an example of a BKB. Section 3 discusses how genetic algorithms can be used to perform belief revision, and also reveals the problems associated using genetic algorithms on BKBs. Section 4 presents a method to evaluate incompleteness, and Section 5 describes our test methodology. Finally, Section 6 presents our conclusions.



Brett Borghetti
Sat Apr 20 16:13:58 EDT 1996