摘要： 多种群遗传算法（MPGA）搜寻最优解的能力受初始种群分布的影响，在解决复杂函数优化问题时存在早熟收敛风险，而思维进化算法（MEA）存在局部搜索精度低和全局收敛速度慢的问题。针对两者的不足，提出一种MPGA和MEA混合的优化算法MPGA-MEA。为参与MEA趋同操作的各子群体设置不同的控制参数，独立进行遗传搜索，同时利用移民算子增强子群体的互动，实现协同进化，直至子群体成熟。在此基础上，释放劣质子群体，并选择全局公告板中记录的优质个体执行交叉和变异操作，产生中心个体，对应生成的临时子群体参与新一轮的迭代寻优。基于不同测试函数的仿真结果表明，该混合算法相较于MPGA和MEA，MPGA-MEA对高维多峰函数的寻优能力得到明显提升。

关键词: 多种群遗传算法, 思维进化算法, 选择操作, 交叉操作, 变异操作, 移民算子, 人工选择算子

Abstract: The ability of the Multiple Population Genetic Algorithm(MPGA) to search for the optimal solution is greatly affected by the initial population distribution, and it has a high risk of premature convergence when solving the problem of complex function optimization.The Mind Evolution Algorithm(MEA) is limited in the local search accuracy and global convergence speed.To address the problems, an optimization algorithm that combines MPGA and MEA is proposed.It sets different control parameters for each subgroup participating in convergence operation in MEA to perform genetic search independently.At the same time, it uses the immigration operator to enhance the interactions between subgroups to realize coevolution until the subgroups mature.On this basis, the inferior subgroups of MPGA-MEA are released, and superior individuals recorded in the global bulletin board are selected to perform crossover and mutation operations to generate central individuals, and the corresponding temporary subgroups are involved in the new round of iterative optimization.The simulation results of several test functions show that compared with the original MEA and MPGA algorithms, MPGA-MEA displays a significant improvement in the optimization ability of high-dimensional and multimodal functions.

Key words: Multiple Population Genetic Algorithm(MPGA), Mind Evolution Algorithm(MEA), selection operation, crossover operation, mutation operation, immigration operator, artificial selection operator

中图分类号: 

• TP399