Probabilistic simplified fuzzy ARTMAP(PSFAM)

Abstract

The probabilistic simplified fuzzy ARTMAP (PSFAM) has been developed for fast training and offline or online learning and classification of data together with a probability measure of confidence in the classification. A simplified fuzzy ARTMAP (SFAM) and a Bayes' classifier are combined. Using a committee of SFAMs and brain-evoked response data from four groups of subjects, classification accuracies in the range 87–97% are achieved together with ideal or near-ideal medical statistics. The limitations appeared to be associated with the data.