Configuration Control of Adaptive Diagnostic Expert Systems

Adaptive expert system configurations are modified as new rules or facts are learned. This is similar to uncontrolled modifications of non-adaptive systems, where the software can be changed without documentation or approval. This problem is intensified in adaptive diagnostic expert systems, where so many of the learned facts or rules are environmentally dependent. This paper addresses some of the conditions causing the varied configurations of these systems and presents some suggested software structures to provide the necessary control.