Differential Evolution Algorithm Based on Coupling-Coordination-Based Mutation Strategy

The differential evolution (DE) algorithm is a heuristic, stochastic, parallel search algorithm. The mutation operation is an integral part of the DE algorithm, relating to the basis and difference among vectors. Recently, many improved variations of mutation strategies have been proposed, and promising results have been achieved. However, under modifications related to the difference vector, individuals are selected for the difference vector mainly based on fitness values, which might decrease the population diversity and affect the algorithm performance. This paper proposes a coupling-coordination-based mutation strategy for the DE (in short for CCDM) to improve the selection of individuals in the difference vector. First, the coupling-coordination degree, which comprehensively considers individuals’ fitness values and distribution, is used to determine similarity between individuals. Then, the population individuals are clustered into four subpopulations according to their similarity. The subpopulation that contains the basis vector individuals serves as a similarity archive for the last vector in the difference vector. Finally, the concept of quantile is used to construct the elite archive for the first vector in the difference vector to accelerate the convergence. The effectiveness of the CCDM is verified through numerical experiments on the CEC2017 test function set using different types of mutation strategies and DE variants. Compared with the existing difference vector improvement strategies, the CCDM can further enhance searchability and convergence.