A PARAMETER FREE FILLED FUNCTION METHOD FOR GLOBAL OPTIMIZATION

The filled function method is an effective approach for global optimization. However, the existing filled functions have the disadvantages of discontinuity, non-differentiability and sensitivity to parameters, which limit their efficiency. In this paper, a parameter free filled function is proposed with two advantages: 1) continuity and differentiability, and 2) no need for adjusting parameters. Then a new filled function algorithm is designed based on the proposed filled function for global optimization problems. Furthermore, to make the filled function algorithm more effective, a new search method with restarting mechanism is designed. Finally, numerical experiments are conducted and comparisons are made with two representative filled function algorithms. The results show the proposed algorithm can find higher quality solutions and has faster convergence speed.