Evolutionary Programming Based on Ladder-changed Mutation for Adaptive System Recognition

Premature convergence is the fatal shortcoming of traditional evolutionary programming. Based on analysis of premature convergence of traditional evolutionary programming, a novel Evolutionary Programming which fast optimize the search apace is proposed. The ladder attenuated mutation operator is introduced in this new EP algorithm to overcome the shortcoming which the individual easy to get into the local minimum in general EP. Simulations confirm this new EP algorithm is better than classic evolutionary programming algorithm in the aspects of global optimization, convergence speed and the robustness. In this paper, system recognition is an example. In order to recognize the unknown system, the adaptive FIR Filter is used to simulate the unknown system, and we use this new Evolutionary Programming Algorithm in this adaptive filter. This algorithm is not depended on any parameters. We can get a good result by the simulation, it indicate the validity and engineering value of the algorithm in signal process