A Wavelet Neural Networks Based on the Genetic Algorithm
Zhang Li · Acta Simulata Systematica Sinica · 2001
One of the important problems in the research of the artificial neural networks is the optimum learning method.Combining the performance of global optimum searching of the genetic algorithm with the time-frequency localization of the wavelet analysis, a new type of the neural networks-- wavelet neural networks of genetic algorithm (GAWNN) is proposed in this paper. The experimental results, which are the XOR problem and the two spirals problem, have demonstrated that the GAWNN not only has the local property of the wavelet analysis and the generalization capability of the artificial neural networks, but also has the advantages of the fast global searching of the genetic algorithm. So the WNNGA can be treated as an ideal algorithm for the training of the multi-layer feeding forward neural networks.