Maintenance in Probabilistic Knowledge-Based Systems.
Abstract
Recent developments using directed acyclical graphs (i.e., influence diagrams and Bayesian networks) for knowledge representation have lessened the problems of using probability in knowledge-based systems. Little has been done concerning the maintenance of domain-specific knowledge, which includes the probabilistic information about the problem domain. By making use of conditional independencies represented in the graphs, probability assessments are required only for certain variables when the knowledge base is updated. This study was investigated, for those variables which require probability assessments, ways to reduce the amount of new knowledge required from the expert when updating probabilistic information in a probabilistic knowledge-based system. Three special cases (ignored outcome, split outcome, and assumed constant outcome) were identified under which many of the original probabilities (those already in the knowledge-base) do not need to be reassessed when maintenance is required. Although some reduction in the number of probability assessments can be achieved when the special cases apply, it appears other areas may be more productive in reducing the level of effort needed to maintain probabilistic KBS's. Topics recommended for future research include the development of efficient propagation techniques for multiply connected graphs, and investigation of methods to make the probability encoding process more efficient.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1987
- Accession Number
- ADA189498
Entities
People
- Thomas F. Reid
Organizations
- Air Force Institute of Technology