一种改进的快速高效的差分进化算法

文章针对差分进化算法收敛速度和全局搜索能力之间不能同时兼顾这一问题,提出了一种改进的差分进化算法,该算法从动态更新种群、递增策略的交叉概率因子及递减策略的缩放因子对标准DE算法进行了改进,并用6个典型的测试函数对改进的差分进化算法和标准差分进化算法进行测试比较,结果表明改进后的差分进化算法在收敛速度、收敛精度和算法鲁棒性方面都要优于标准差分进化算法,采用动态更新种群的策略也有效地提高了算法的运算效率.

A modified fast and highly efficient differential evolution algorithm

In order to solve the contradiction between the velocity of convergence and the ability of global optimization in the differential evolution(DE) algorithm, a modified DE(MDE) algorithm is presented,in which dynamic updating of the population, increasing of the crossover factor and decreasing of the scaling factor with the generation are considered. Six typical test functions are adopted to make a comparison with the standard DE algorithm. The experimental results show that the MDE algorithm is superior to the DE in velocity of convergence, precision of optimization and the robustness. Moreover, the method of dynamic updating population can increase the computation efficiency of the algorithm.