Proposal of Adaptive T-S Fuzzy Radial Basis Function Network

Aiming at the parameter identification of T-S fuzzy inference system,an adaptive T-S fuzzy radial basis function neural network was proposed by using the knowledge representing property of fuzzy inference system and the self-learning property of neural network fully. In order to design a minimum structure network with expected accuracy,a structure-learning algorithm to construct the network structure dynamically was given based on the analysis of the learning dynamics of network. The hidden units could be added,merged and deleted in an adaptive way according to the task complexity and the progress of learning without any prior knowledge. When this fuzzy RBF network is applied to nonlinear function approximation,a good result is obtained.