Supervised ART-I: A new neural network architecture for learning and classifying multi-valued input patterns

Abstract

A new supervised neural network architecture has been introduced, called Supervised ART-I. It has the accuracy of Fuzzy ARTMAP in classifying of both binary and analog arbitrary multi-valued input patterns. However, it is quicker in learning and classifying, has fewer parameters and requires less memory, due to its simple architecture. The Supervised ART-I has been built from a single Fuzzy ART, instead of a pair of them as in Fuzzy ARTMAP. During the training phase, when the tag of the winning node matches the class code, the node will be trained. If not, class correction is done. If non of the committed nodes is able to represent the input, a new node is committed and it is tagged with the current class code. During the testing phase, if the winning node passes the vigilance parameter (p), its tag represents the class of the input, if not the network fails to classify the current input. The architecture, learning, and testing of the network have been discussed. The full algorithm has been listed.