Neuro-fuzzy models in pattern recognition

Research in the last decade emphasized the design of adaptive pattern recognition classifiers using multi-layered artificial neural nets. The greatest potential in this was not only in the adaptivity but also in the high-speed processing through massively parallel VLSI implementation and optical computing. Computational advantages of such algorithms have been discussed in a number of papers. Neural networks, particularly the self-organizing types, are suitable for crisp pattern clustering of unlabeled data sets. The generalization of Kohonen-type learning vector quantization (LVQ) clustering algorithm to fuzzy LVQ (FLVQ) clustering algorithm and its equivalence to fuzzy c-means have been presented recently. On the other hand, Carpenter/Grossberg's adaptive resonance theory (ART) has been modified to perform fuzzy clustering by a number of researchers. The performance of these neuro-fuzzy models in clustering unlabeled data patterns is addressed in this paper. A recent development of a new similarity measure and a new learning rule for updating the centroid of the winning cluster in a fuzzy ART-type neural network is also described. We demonstrate the capability of this new algorithm to better represent the nonlinear decision boundaries of neighboring clusters embedded in unlabeled data sets using computer generated data.