Research of Optimal Parameter Settings for Improved Wavelet Neural Network

The deficiency in parameter settings for improved wavelet neural network based on ant colony algorithm is analyzed furthermore in this paper and the optimization condition of key parameters in wavelet network based on the orthogonal design is proposed as well. Under the balance-col location theory of the orthogonal design, the relation of primary and secondary of network output properties which is effected by the function of various parameters with interaction each other in the process of learning algorithm of wavelet network is analyzed during the design for the test operation, and the optimum combination scheme of wavelet network parameter is confirmed in the same time. The result of simulation experiment shows that the times of parameter selection are decreased, the period of experiment is shortened, and the subjective problem in the process of parameter selection of wavelet network is avoided efficiently by the scheme of parameter optimization which is discussed in this paper. However, how to further optimize the structure of wavelet network further to realize the adaptive estimation of parameters still needs further research.