LEARNING AND GENERALIZATION OF NOISY MAPPINGS USING A MODIFIED PROBART NEURAL-NETWORK

Incremental function approximation using the PROBART neural network of fers many advantages over conventional feedforward networks, These inc lude dynamic node allocation based on the complexity of the function a pproximation task, guaranteed convergence, and the ability to handle n oise in the training data, However, the PROBART network does not gener alize very well to untrained data, In this paper, a modified PROBART i s proposed to overcome this deficiency, This modification replaces the winner-take-all mode of prediction of the PROBART with a distributed mode of prediction, This distributed mode enables several neurons to c ooperate during prediction and, thus, provides better generalization c apabilities even in noisy conditions, Computer simulations are conduct ed to evaluate the performance of the modified PROBART neural network using three benchmark nonlinear function approximation tasks, The pred iction accuracy of the modified PROBART network compares favorably to the PROBART, fuzzy ARTMAP, and ART-EMAP networks for all these tasks.