Blind Equalization by Wavelet Neural Network with Nonlinear Memory Gradient Algorithm

Blind equalization by wavelet neural network with nonlinear memory gradient algorithm was proposed in this paper. The traditional blind equalization by wavelet neural networks using stochastic gradient algorithm to update the network weights, which has shortcomings of slow convergence rate and easy to fall into local minimum. To further improve the performance of blind equalization by wavelet neural network, memory gradient algorithm is adopted to update the network weights. Memory gradient algorithm can effectively use the current and previous iterative gradient information and then has fast convergence rate and to some extent to avoid falling into local minima. because the objective function of blind equalization by wavelet neural network is non-convex and noise interference in the practical communication system, memory gradient algorithm for linear search can not guarantee the downward direction of gradient after each iteration, therefore, nonlinear transformation for gradient function is carried out to maintain the stability and has faster convergence rate at the same time. Computer simulation and pool experiment results show that blind equalization by wavelet neural network with nonlinear memory gradient algorithm has better performance compare with feed-forward BP network and traditional wavelet neural network.