Dynamic fuzzy neural networksa novel approach to function approximation.pdfVIP

358 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B: CYBERNETICS, VOL. 30, NO. 2, APRIL 2000 Dynamic Fuzzy Neural Networks—A Novel Approach to Function Approximation Shiqian Wu and Meng Joo Er Abstract—In this paper, an architecture of dynamic fuzzy neural networks (D-FNN) implementing Takagi–Sugeno–Kang (TSK) fuzzy systems based on extended radial basis function (RBF) neural networks is proposed. A novel learning algorithm based on D-FNN is also presented. The salient characteristics of the algorithm are: 1) hierarchical on-line self-organizing learning is used; 2) neurons can be recruited or deleted dynamically according to their significance to the system’s performance; and 3) fast learning speed can be achieved. Simulation studies and com- prehensive comparisons with some other learning algorithms demonstrate that a more compact structure with higher performance can be achieved Fig. 1. Architecture of the D-FNN. by the proposed approach. Index Terms—Dynamic structure, function approximation, fuzzy neural networks, hierarchical on-line self-organizing learning, TSK fuzzy inference system (ANFIS) is mainly determined by expert knowledge. reasoning. In [4], the author proposed the self-organizing learning scheme for structure learning. A hierarchically self-organizing approach was used in [5]. In [6], the structure is determined by clustering the input I. INTRODUCTION