一种融合邻域搜索的多策略差分进化算法

设计多策略差分进化算法的难点在于选择何种变异策略以及如何分配这些策略。提出一种融合邻域搜索的多策略差分进化算法，根据个体适应度值将种群分为3 个子种群，每个子种群分别采用不同的变异策略和参数值，使得各子种群的搜索能力可互补，有助于平衡整个种群的勘探和开采能力。同时，对适应度值最好的子种群采用邻域搜索操作，充分挖掘优质个体可能包含的有益信息用于指导搜索。在34 个测试函数上实验，与包含7 种差分进化算法在内的12 种进化算法进行对比，结果表明该算法在大多数函数上取得了更好性能。

A Multi-strategy Differential Evolution Algorithm Combined with Neighborhood Search

The difficulties of designing a multi-strategy differential evolution (DE) algorithm are how to select the mutation strategies and allocate these strategies. A multi-strategy DE algorithm combined with the neighborhood search operator is proposed. The population is divided into three subpopulations according to the fitness values, and each subpopulation employs a different mutation strategy and parameter settings to complement the search ability, to balance the exploration and exploitation ability of the whole population. The subpopulation with the best fitness values employs the neighborhood search operator to exploit possible benefit information to guide the search. Extensive experiments are carried out on 34 test functions to compare with 12 different evolutionary algorithms, which include the 7 DE algorithms. The results show that the algorithm can perform better on most test functions.