A new approach to generate a self-organizing fuzzy neural network model
Guannan Leng, Girijesh Prasad, T.A. McGinnity · 2003
This paper presents a new approach for creating a self-organizing fuzzy neural network (SOFNN) from training data, to implement the Takagi-Sugeno-Kang (TSK) model. The center vector and the width vector have been introduced in the RBF neurons in the SOFNN. Novel methods of structure learning and parameter learning, based on new adding and pruning techniques and a recursive on-line learning algorithm, are proposed and developed. The proposed methods are very simple and effective and generate a fuzzy neural model with a high accuracy and a very compact structure. Simulation studies based on a pH neutralization process, confirm that the SOFNN has the capability of self-organization, and can determine the structure and parameters of the network automatically without non-linear optimization.