Wavelet neural network based on sampling theory for non uniform noisy data

Training wavelet neural network based on sampling theory has been shown to have global convergence and avoid overfitting. In this paper we improve this approach for constructing and training the wavelet network using non uniform and noisy data. We first propose a method to find the appropriate feedback matrix for training the wavelet network. Then we use early stopping and wavelet thresholding to optimize the wavelet network structure and overcome the overfitting problem. Performance of the proposed method has been tested on one and two-dimensional functions. The presented results demonstrate the effectiveness of the proposed methods to decrease the generalization error in training non uniform and noisy data and also the reduction of the complexity of the wavelet network.