A novel learning algorithm for dynamic fuzzy neural networks
Shiqian Wu, Meng Joo Er, Jun Liao · 1999
A learning algorithm for dynamic fuzzy neural networks based on extended radial basis function (RBF) neural networks, which are functionally equivalent to TSK fuzzy systems, is proposed. The algorithm comprises four parts: (1) criteria of neurons generation; (2) allocation of parameters of RBF units; (3) weight adjustment; and (4) pruning technology. The algorithm has fast learning speed as the weights are modified by the linear least square method and no iteration is needed. The synergy of fuzzy and neural systems with dynamic structure shows that a parsimonious structure with high performance can be achieved. Simulation studies and comprehensive comparisons with some other learning algorithms demonstrate that the proposed algorithm is superior. It is also shown that the method is promising for real-time applications.