1. A belief management architecture for diagnostic problem solving
- Author
-
Gautam Biswas, Benoit M. Dawant, Kazuhiko Kawamura, and Serdar Uckun
- Subjects
Computer science ,business.industry ,Frame (networking) ,Control (management) ,Principal (computer security) ,Space (commercial competition) ,Machine learning ,computer.software_genre ,Discernment ,Artificial intelligence ,Decision-making ,Architecture ,business ,computer ,Abstraction (linguistics) - Abstract
An architecture for diagnosis that uses qualitative endorsements as its principal method of uncertainty abstraction and propagation is presented. The framework performs local belief computations in a hierarchical hypothesis space, in contrast with methods that propagate evidence throughout the whole frame of discernment. In this system, global control of the decision making process is maintained by local evaluations of belief status. These local evaluations determine an active focus in which refinement of belief status is undertaken by gathering additional information. The main goal of the research project is the development of a framework for reasoning with endorsements, and the diagnostic application explicated in the paper is built as a proof-of-principle.
- Published
- 1990
- Full Text
- View/download PDF