Differential evolution with migration mechanism and information reutilization for global optimization

Differential evolution (DE) is an efficacious global optimization algorithm, and many variants have been advanced since its inception. During the iterative search process, any individual in any DE-type algorithm can likely locate a local optimum, and once that happens it may need many attempts for this individual to find another better solution, thus leading to ineffective consumption of computing resources and decline in the opportunity to search other promising regions. Therefor, this article proposes a DE with migration mechanism and information reutilization (MIDE). Specifically, a migration mechanism is first presented to make individuals located at local optima abandon current locations and move to other regions to continue their search, tending to solve the problem above. Meanwhile, a new mutation strategy named DE/pbest/1 with external archive is introduced to reutilize abandoned local optima to provide helpful information of evolution. Additionally, the settings of control parameters associated with this mutation strategy are designed in such a manner that they can contribute to well balancing exploration and exploitation. To evaluate the performance of MIDE, extensive experiments are conducted on CEC 2017 and CEC 2014 test suites, and the comparison results between MIDE and 19 competitors (including 13 state-of-the-art DE variants and six winner algorithms of CEC 2014 and CEC 2017 competitions) demonstrate MIDE's competitive performance.