RJADE/TA Integrated with Local Search for Continuous Nonlinear Optimization

This paper proposes a memetic algorithm by integrating adaptively a local search approach with a recently proposed variant of differential evolution, reflected adaptive differential evolution with two external archives (RJADE-TA). The main objective is to bring together the exploration factor of differential evolution and exploitative component of local search to solve continuous optimization problems. A novel hybrid local search mechanism is proposed and demonstrated leading to a crossbred version of RJADE-TA. In other words, the best solutions after a regular toll of global search are migrated to an archive, where Davidon Fletcher Powell local search method is implemented to the migrated solutions. Afterwards, the population is updated with new reflected solutions to prevent premature convergence. The proposed approach is novel in the sense that most of the algorithms store only inferior or superior solutions in the archives. None of the algorithms implemented the local search inside the archive. Thus, this combination is a new one. To evaluate the merit of developed meme, a benchmark suite of complex 28 functions from CEC 2013 test problems is selected and implemented. The experimental results demonstrate that this integration of local search strategy can further improve the performance of RJADE-TA. They further reveal that the proposed meme outperforms differential evolution based algorithms on most of the tested problems.