A COMPARATIVE STUDY OF NEURALNETWORK MODELS

The recent proliferation of neural network models for pattern recognition and other applications has led to a need for benchmarking and a comparative study of such models in order to assist in the development of new models more pertinent to such specific applications. This paper uresents a comoarative studv of the followine three imnortant tvoes of neural network models: (i) ihe Bidirectional Associa&e Memory (BAM), (ii) ihe Back-Propagation (BP), and (iii) the Adaptive Resonance Theory (ART). The models have been analyzed to establish their efficiency, accuracv. and adaotabilitv. Simulation of a unified comnuter imolementation of the models ( the ,. 1 I same language, the same computer, and the same benchmark) can reveal the essential differences in the performance of the models, rather than the differences due to their implementations.