An efficient implementation of a learning method for Mamdani fuzzy models

Abstract

This paper presents an efficient implementation of a supervised learning method based on membership function training in the context of Mamdani fuzzy models. The main idea is to adjust the antecedent and consequent membership functions that are of asymmetric trapezoidal form by backpropagating the output error through the fuzzy net. The proposed implementation is analogous to the training scheme commonly used with Takagi-Sugeno fuzzy models but it requires additional procedures that are related to some specific characteristics of the Mamdani fuzzy structures. Some numerical results are provided as illustrations.