Related Experiment Videos
Specification of models in large expert systems based on causal probabilistic networks
1Department of Medical Informatics and Image Analysis, Aalborg University, Denmark.
Artificial Intelligence in Medicine
|June 1, 1993
Summary
Specifying large expert systems is simplified by using models for conditional probability tables in causal probabilistic networks. A new header facility reduces complexity and improves maintenance of these networks.
Area of Science:
- Artificial Intelligence
- Computer Science
Background:
- Expert systems require detailed specifications, particularly causal probabilistic networks (CPNs).
- CPNs necessitate conditional probability tables (CPTs) for all nodes, which can be complex to define.
Purpose of the Study:
- To address challenges in specifying large causal probabilistic networks.
- To introduce a method for simplifying the definition and maintenance of CPTs.
Main Methods:
- Describing various models for specifying conditional probability tables.
- Presenting a program designed to handle these models.
- Proposing a header facility for common definitions of repeated elements in CPNs.
Main Results:
- Models can effectively describe interactions between nodes in CPTs.
- The proposed header facility significantly shortens specifications.
- Specifications become easier to construct and maintain.
Conclusions:
- The described models and header facility offer a practical solution for specifying large CPNs.
- Implementing these methods enhances the efficiency and usability of expert systems.