This paper presents a new learning method to improve noise tolerance in Fuzzy ART. The two weight vectors: the top-down weight vector and the bottom-up weight vector are differently updated by a weighted sum and a fuzzy AND operation. This method effectively resolves the category proliferation problem without increasing the training epochs in noisy environments.

An effective learning method for the fuzzy ARTMAP in the recognition of noisy input patterns is presented. the weight vectors of the system are updated using the weighted average of the noisy input vector and the weight vector itself. This method leads to stable learning and prevents the excessive update of the weight vectors which may cause performance degradation. Simulation results show that the proposed method not only reduces the generation of spurious categories, but aloso increases the recognition ratio in the noisy environment.