An Improved Differential Evolution with a Novel Restart Mechanism
Mengnan Tian, Xingbao Gao, Xueqing Yan · 2016
This paper presents a novel differential evolution (DE) algorithm for solving global optimization problems by developing a combined local mutation strategy and a restart mechanism. To alleviate the premature convergence and stagnation, a combined local mutation strategy is first developed to improve the exploitation of DE by using two local mutation strategies. Then, a new restart mechanism is proposed to enhance the population diversity and exploit the useful information of superior individuals by searching the super-rectangle using superior individuals, and replacing the inferior individuals with a probability by the ones randomly generated from search space. Furthermore, a simple and efficient approach is applied to adjust control parameters. Finally, the proposed algorithm is compared with four DE variants on 14 well-known benchmark functions. The experimental results show that the proposed approach is very competitive.