NETWORK PROCESSING OF HIERARCHICAL KNOWLEDGE FOR CLASSIFICATION AND DIAGNOSIS.

A system is described that uses a parallel network-based approach to do classification and diagnosis. The principles used in compiling and processing a network are presented. The mechanism used by activation networks that model human perceptual, comprehension, memory, and motor processing is borrowed and applied to the task of diagnosis. Although the network organization and processing that the system utilizes do not provide a psychologically valid model for such a 'higher level' cognitive task that requires conscious attentional control, they do provide a way to use connectionist processing for diagnosis.