Introduction

Adaptive Resonance Theory (ART) was developed by Stephen Grossberg in his studies of the behaviour of models of systems of neurons. Grossberg has published a large number of papers, in which he develops the mathematical theory of such models, usually in terms of systems of differential equations, and applies this to actual neural systems. Like Kohonen, he has been particularly interested in systems that are capable of organizing themselves.

Grossberg has used his mathematical results to develop two neural network architectures, which implement nearest neighbour pattern classification. The first architecture, usually referred to as ART1, works on patterns of binary data. The second, ART2, is an extension of ART1 to arbitrary patterns. ART1 is described in [8], ART2 in [9] and [10].