Geometrical-based category choice fuzzy art architecture

Abstract

In this paper, the category choice of the Fuzzy adaptive resonance theory (ART) is shown to be replaced by a distance measure related to the L-1 norm. This replacement has several advantages. One advantage is that the new distance measure will operate directly on the input patterns without the need for doing complement coding which is a requirement for the category choice of the Fuzzy ART. Another advantage is that using the new distance measure the input patterns do not have to be normalized to be clustered. It is noted that using Fuzzy ART the input patterns need to be in the interval [0, 1]. The difference between the two distance measures is illustrated mathematically and geometrically. Simulation results on different databases are presented.