An Adaptive Differential Evolution Algorithm Based on New Diversity

A DE approach based on a new measure of population diversity and a novel parameter control mechanism is proposed with the aim of introducing a good behavior of the algorithm. The ratio of the new defined population diversity of different generations is equal to that of the population variance, therefore the adaption of parameter can use some theoretical results in19. Combining with the method in18, we can adjust the mutation factor F and the crossover rate CR at each generation in the searching process. The performance of the proposed algorithm (DE-F&CR) is compared to the basic DE and other four DE algorithms over 25 standard numerical benchmarks provided by the IEEE Congress on Evolutionary Computation 2005 special session on real parameter optimization. The results and its statistical analysis show that the DE-F&CR generally outperforms the other algorithms in multi-modal optimization.