Abstract
The applicability of PDP (parallel distributed processing) models to knowledge processing is clarified. The authors evaluate the diagnostic capabilities of a prototype medical diagnostic expert system based on a multilayer network. After having been trained on only 300 patients, the prototype system shows diagnostic capabilities almost equivalent to those of a symbolic expert system. Symbolic knowledge is extracted from what the multilayer network has learned. The extracted knowledge is compared with doctors' knowledge. Moreover, a method to extract rules from the network and usage of the rules in a confirmation process are proposed.

This publication has 5 references indexed in Scilit: