A fuzzy ARTMAP model with contraction procedure

This paper presents a neural network classifier based on fuzzy ARTMAP with conflict-resolving strategy. The proposed model explicitly resolves overlaps among prototypes of different classes through deploying a contraction procedure in the network, therefore, improving its generalization. Compared with other existing methods, the model has the priority of intuition and no parameter tuning. The performance of the model is evaluated using several benchmark data sets. The comparisons with original fuzzy ARTMAP model and other classifiers indicate that the proposed classifier achieves competitive performance.