A METHOD OF DYNAMICALLY TUNING BP ALGORITHM PARAMETERS ACCORDING TO FUZZY RULES

An analysis is made first of a standard BP algorithm,and then an improved method is proposed,which is meant to dynamically adjust learning rate and momentum factor by means of fuzzy inference.The fuzzy control of BP algorithm is thus achieved in the fuzzy inference system.Finally,examples are used to compare this fuzzy controlled BP algorithm with the standard BP algorithm and Vogl’s improved algorithm.The experimental results indicate that it is a method of great promise to improve the performance of BP algorithm with the help of fuzzy inference.