Fuzzy wavelet networks for function learning

Inspired by the theory of multiresolution analysis (MRA) of wavelet transfo rms and fuzzy concepts, a fuzzy wavelet network (FWN) is proposed for appro ximating arbitrary nonlinear functions in this paper The FWN consists of a set of fuzzy rules. Each rule corresponding to a sub-wavelet neural network (WNN) consists of single-scaling wavelets. Through efficient bases selecti on, the dimension of the approximated function does not cause the bottlenec k for constructing FWN, Especially, by learning the translation parameters of the wavelets and adjusting the shape of membership functions, the model accuracy and the generalization capability of the FWN can be remarkably imp roved. Furthermore, an algorithm for constructing and training the fuzzy wa velet networks is proposed. Simulation examples are also given to illustrat e the effectiveness of the method.