A Learning Algorithm for Feedforward Neural Networks Based on Fuzzy Controller

Abstract

This paper proposes a continuous learning algorithm for feedforward neural networks, an improvement of traditional BP algorithm. Then analyze its convergence and bring up one fuzzy control algorithm to realize adjusting control variables adaptively in order to improve its performance in convergence. At last, we apply this continuous learning algorithm based on fuzzy controller to lateral prediction for reservoir parameters, which make good performance in error control and convergence. This indicates that the algorithm has significance in practice.