Toward Data Analysis Using Connectionist Modules.

A connectionist model for estimating the posterior probability of disease is presented. It is constructed of several back propagation modules, each of which serves a particular function. Manifestations serve as input to the network; they may be real-valued, and the confidence in their measurement may be specified. The network produces as its output the posterior probability of disease. In the tests that were performed, the calculated error was 0.025.