Tuning selection pressure in differential evolution using local selection

Population structuring is used in cellular evolutionary algorithms (cEAs) as an effective means of controlling the selection pressure and thereby the exploration/exploitation tradeoff of the algorithm. Traditionally the selection intensity in differential evolution (DE) algorithm has been tuned using its controlling parameters. In this work, we investigate the possibility of adjusting the selection pressure in DE using local selection. Studying with different neighborhood structures, it was found that the local selection mechanism offers plenty of promises in tuning the selection pressure of DE. The experimental results reveal that the cellular DE (cDE) algorithm performs better sampling of the search space and thereby improve the convergence characteristics of the algorithm compared to the canonical version.