Co-evolutionary differential evolution with dynamic population size and adaptive migration strategy
Zhi‐Hui Zhan, Jun Zhang · 2011
As the performance of differential evolution (DE) is significantly affected by its mutation schemes and parameter settings when solving different problems, this paper proposes a simple yet efficient co-evolutionary DE (CEDE) to enhance the algorithm performance. The CEDE algorithm uses multiple populations to optimize the problem cooperatively, with each population using different operators and/or different parameters. Moreover, as different populations may show different performance on the same problem, we further design an efficient adaptive migration strategy (AMS) to dynamically control the population size of different populations. The CEDE algorithm is tested and compared on four benchmark functions. Experimental results demonstrate the good performance of CEDE when compared with conventional DEs using different operators and/or parameters.