Abstract
So far, the competitive associative net called CAN2 has been developed to utilize the competitive and associative schemes for learning to achieve efficient piecewise linear approximation of nonlinear functions. Although the conventional online learning methods for the CAN2 have been shown effective, they basically are for infinite number of training data. Provided that only a finite number of training data are given, however, the batch learning scheme seems more suitable. We here present a batch learning method to learn a finite number of training data efficiently by means of combining competitive learning, associative learning and reinitialization using asymptotic optimality. Finally, we apply the present method to learning to approximate sevaral artificial benchmark functions and show that the batch CAN2 calculates faster and achieves smaller MSE (mean square error) than the conventional online CAN2, and it has several advantages superior to the SVR (support vector regression).
