An enhanced neural network learning using a self-tuning fuzzy neuron controller

Abstract

The performance of the "fuzzy-controlled curved search" method for back propagation neural network learning depends heavily upon the membership functions used in the fuzzy controller. Manually tuning the membership functions becomes a tedious job. In this paper, a fuzzy neuron controller with self-tuning capability is introduced to adjust the related membership functions in a self-adaptive manner. Computational results are included to illustrate the potential of this enhanced learning method.