A new learning method to improve the category proliferation problem in fuzzy ART

This paper presents a new learning method to solve the category proliferation problem in fuzzy ART. In the conventional learning methods, both the top-down and bottom-up weight vectors are updated by the fuzzy AND operation between the input vector and the weight vector (or template). But in the proposed method, these vectors are differently updated: the top-down weight vector is updated by the weighted sum of both its current value and the input vector, and the bottom-up weight vector is updated by the fuzzy AND operation between its current value and the newly learned top-down weight vector. The proposed learning method can prevent the abrupt change of the template and achieve good noise tolerance in noisy input pattern. Simulation results show that the proposed learning method effectively resolves the category proliferation problem without increasing the training epochs in noisy environments