Radial-basis-function neural network based on fast recursive algorithm and its application
基于快速回归算法的RBF神经网络及其应用

Considering the difficulty in selecting the numbers and determining the locations of the centers of radial basis functions (RBF) in the RBF neural network (RBFNN), a novel RBFNN is proposed based on the fast recursive algorithm (FRA). Using FRA, we can determine the numbers and locations of the centers, and derive the weights between the hidden layer and the output layer. The new RBFNN is used to fit a single-variable function curve and predict the Mackey-Glass chaotic time series. The simulation results demonstrate the effectiveness and practicability.