A Historical Information Based Differential Evolution
Yifan Qin, Libao Deng, Chunlei Li, Wenyin Gong · 2023
Differential evolution is an efficient and robust optimizer. However, DE still has the problems of evolutionary stagnation and inappropriate generated control parameters. In the search behavior of each individual in the past population, some valuable historical information for future evolution may be hidden, which can help the optimizer solve these problems. Based on the above consideration, we propose a historical information based differential evolution (HIDE). In our algorithm, a new mechanism is established to judge whether an individual is in stagnation, and a new mutation strategy based on discarded parent vectors from different periods is proposed to help the stagnant individuals escape from the local optimum. Meanwhile, we designed a new update method for control parameters based on historical information. Compared with the mainstream schemes, the parameters generated by our method are more suitable for the current function. To evaluate the performance of our algorithm, we compared HIDE with eight advanced variants on the CEC 2017 test platform. The experimental results show that the quality of the solution provided by HIDE is better than that of other variants.