Research on Regularization Network Optimization Based on Wavelet Function

In order to solve overfitting of modeling in noisy circumstance, a normal function with corresponding training algorithm is proposed for wavelet networks based on wavelet sampling theory. Since such an algorithm can use sample distributions and errors respectively to train input and output weights, learning efficiencies of wavelet networks are improved greatly. In addition, the related theorems have also been given to demonstrate that the novel cost function can ensure minimization of the approximation error, the theories and experiments show that this novel cost function can ensure generalizations of wavelet networks. Data management module is responsible for data communication in the test process. It can obtain data from external or instrument drive module and transmit data to signal processing module for processing. When completion of the test, the test results are edited and printed through the affiliated function modules.