ART1 : model algorithm characterization and alternative similarity metric for the novelty detector

This paper investigates adaptive resonance theory 1 (ART1) as a pattern clustering algorithm. Inherent characteristics of the model are analyzed. In particular the vigilance parameter, p, and its role in classification of patterns is examined. The authors´ experiments show that the vigilance parameter as designed originally does not necessarily increase the number of categories with its value but can decrease also. This is against the claim of solving the stability-plasticity dilemma. However, the authors have proposed a modified vigilance test criterion which takes into account the problem of subset and superset patterns and stably categorizes arbitrarily many input patterns in one list presentation when the vigilance parameter is closer to one. The proposed method also performs much better with regard to classification of patterns and reduces the number of list presentations required for stable category learning. Improved performance of the new similarity criterion with regard to noisy patterns has been demonstrated experimentally