A NOVEL DIFFERENTIAL EVOLUTION USING MULTIPLE-DEME BASED MUTATION

作者：Lim Dong Hyun; Luong Hoang N; Ahn Chang Wook^*

来源：International Journal of Innovative Computing Information and Control, 2012, 8(5A): 3049-3060.

• Differential evolution
• Evolutionary algorithm
• Numerical optimization
• Population diversity
• Parallel processing
• Multiple-deme based mutation

Mutation strategy has been acknowledged to significantly influence the performance of differential evolution (DE). The popular DE/rand/1 can relatively maintain the diversity of population, but the algorithm often distributes its search effort widely among search space without any focus. On the other hand, greedy strategies, such as DE/best/1 or DE/current-to-best/1, have faster convergence speed, but they are likely to fall into local optima due to insufficient population diversity. In this paper, we propose a novel DE variant which is able to sustain the balance between exploration and exploitation. We introduce the concept of parallel computing into DE and design a multiple-deme based mutation operator (MDM) to enable information exchange among processing units. Experimental results over a number of numerical optimization problems prove that our proposed differential evolution, DE-MDM, outperforms the traditional DE approach in terms of the quality of the final achieved solutions.

• 出版日期2012-5