Complex-valued function approximation using an improved BP learning algorithm for wavelet neural networks

A new complex-valued wavelet neural network is proposed in this paper by introducing an additive momentum and new error function for the complex-valued wavelet neural network, in which the defect of gradient descent of traditional complex-valued back propagation algorithm can be avoided. It is used for the complex-valued function approximation to verify its feasibility and effectiveness. The simulation results show that the new network has better convergence, better stability and faster running speed than the traditional complex-valued wavelet neural network and complex-valued back propagation network.