Abstract: Existing network balancing algorithm has the problems of the slow convergence and the computational redundancy. In response to this phenomenon, through the analysis of traditional adaptive equalization algorithms and wavelet transform, we proposed a network equalization algorithm based on wavelet transform. The robustness of the wavelet transform compensates the slow convergence of the adaptive equalization algorithm. Through analyzing the convergence of the algorithm, the iteration parameters are reset. Experimental results show that the experimental results are basically consistent with the expected results. It is of a good convergence effect and a lower error rate.

Key words: neural networks, balancing algorithm, wavelet transform, convergence, iteration