基于多种群的自适应差分进化算法

在分析了经典和改进变异操作算子的属性以及种群统计信息的基础上,按照个体适应度的差异,将个体分成不同的子种群并针对不同的个体适应度值,采用不同的变异算子,以保证在加快算法收敛速度的同时有效地跳出局部极值点.在参考经验值的基础上,加以自适应调整,使算法达到全局搜索能力与局部搜索能力的平衡.针对13个标准测试函数的仿真实验结果表明,所提出的算法与其他算法相比较具有较好的效果.

Adaptive Differential Evolution Algorithm Based on Multiple Subpopulation with Parallel Policy

Analyzing the attributes of classic and modified mutation operators and the statistical information on population, all of the individuals is classified into different subpopulation according to the difference between individual adaptability and different mutation operators are introduced according to the different values of individual adaptability so as to get rid of the local optima efficiently with convergence rate expedited. Based on the empirical values of control parameters, the parameters are adjusted adaptively to balance the global and local search capability in an algorithm. The simulation results of 13 standard test functions showed that the algorithm proposed is more effective in comparison with other algorithms.