Improved Wavelet Networks Algorithm Research and its Application

Abstract

The conventional learning algorithm based on BP method may converge to a local minimum, slowly converging speed and is shock before and after the Convergence point. A algorithm based on BP and PID techniques for wavelet network learning was proposed. And PIDBP algorithm can significantly reduce the probability of the emergence of local minimum after adding momentum term, At the same time introduction of the inertia term, Can be in the larger learning parameters to speed up the convergence and divergence and to reduce the possibility of oscillation, And avoid conventional BP algorithm in the convergence region Insensitivity to accelerate the convergence.