A Connectionist Approach to the Diagnosis of Dementia
Benoit H. Mulsant, Emile Servan-Schreiber · PubMed Central · 1988
This paper describes an implemented connectionist network that performs clinical diagnosis in the domain of dementia. During the past decade, connectionism --also called parallel distributed processing or neural processing-- has been established as a new cognitive and computational paradigm, with strong claims that it provides powerful mechanisms to bring solutions to problems previously intractable. To study the suitability of connectionist networks to perform a sequential diagnostic classification task under uncertainty, we have implemented a network that learns to diagnose cases of dementia. We describe in detail the implementation, training, and behavior of this network. We also discuss directions for future research suggested by the limitations of this network.