A hybrid simplex differential evolution algorithm
Lianghong Wu, Yaonan Wang, Xiaofang Yuan, Shaowu Zhou · 2010
According to the disadvantage of slow convergence rate of the basic differential evolution (DE) algorithm, a hybrid optimization algorithm incorporated Nelder & Mead (NM) simplex method into the basic DE algorithm is presented in this paper. This hybrid procedure performed the exploration with DE and the exploitation with the NM simplex method. Sensitivity to the control parameters of the proposed approach is analyzed. The computational results on several classical Benchmarks nonlinear complex functions show that the hybrid optimization algorithm is superior to the two original search techniques (i.e. NM and DE) in terms of solution quality and convergence rate. Compared with other DE variants, the proposed algorithm has better convergence performance and robustness. The Wilcoxon non-parametric statistical tests also confirm the above claims.