Orthogonal Multiwavelets Neural Network Ensemble and its Application to Structure Approximate Calculation

In this paper, a model of orthogonal multiwavelets neural network ensemble is proposed. The neural network ensemble consists of component orthogonal multiwavelets neural networks where each component neural network is trained by back propagation (BP) algorithm and with orthogonal multiwavelets functions in the hidden layer. Due to the orthogonality of orthogonal multiwavelets functions, all the hidden nodes are orthogonal and all the component neural networks are orthogonal, which can reduce the redundancy and improve the prediction accuracy for the network. The experimental results demonstrate that the proposed neural network ensemble has better generalization performance than BP neural network ensemble.