An automatic control parameters tuning method for differential evolution

Abstract

Differential Evolution(DE) is one of the evolutionary algorithms, and has shown excellent ability for solving global optimization problems over continuous spaces. DE has a few control parameters that have to be set by users. The ability of DE is depending on these control parameters, and, in many cases, trial-and-error process are required for finding suitable parameters. This paper proposes a method for tuning the DE control parameters automatically. The proposed method tunes control parameters within DE search process. By using this method, trial-and-error processes for finding appropriate control parameters are not required. To evaluate the performance, the proposed method was applied to 9 benchmark function optimization problems. The results indicate that the proposed method shows 100% feasibility to solving the optimization problems without outstanding increase of generations required for finding the optimum solutions. © 2008 The Institute of Electrical Engineers of Japan.