In recent years, artificial neural network, whose highly nonlinear, parallel, good fault tolerance and other advantages, in a multidisciplinary study has been widely used. After years of development, although the neural network is a more mature discipline, it has made remarkable progress, in course of the study, we found that some of the pitfalls inherent in neural networks has greatly limited its further development and application. For example, RBF neural network to the existence of prior vector n...