摘要
把SSO算法的交叉策略、协方差矩阵学习策略与传统的DE算法结合,提出一个新的DE算法的变种,我们把它称作SCDE算法。正如我们所知,DE算法的变异策略在DE算法中占据了非常重要的位置,然而,传统的DE算法的变异策略都是用相对位置来产生候选解,本文尝试利用个体历史最优解来诱导变异产生候选解,这将大大提高种群跳出局部最优的能力。此外,将算法的变异和交叉操作放在由种群的协方差矩阵的所有特征向量组成的坐标系中执行,这将使算法的交叉和变异操作具有旋转不变性。实验结果表明,本文提出的新的交叉和变异策略可以大大提高DE算法在CEC 2013中28个测试函数的全局寻优能力。
We propose a new differential evolution (DE) algorithm which incorporates the DE with the crossover thought of the simplified swarm optimization (SSO) algorithm and the covariance matrix learning, called SCDE algorithm. As we know, mutation operation plays a very important role in the performance of the DE. However, traditional mutation strategies of the DE all use the relative position to generate a candidate vector. We try to utilize individual historical optimal solution to induce the variation, which can greatly enhance its ability of jumping out of local populations. Furthermore, the crossover and mutation operations are executed in an appropriate coordinate system which is generated by the covariance matrix of the population and it can make the crossover and mutation operation rotationally invariant. Experimental results show that the proposed algorithm can significantly improve the DE's performance on a set of 28 test functions in CEC 2013 benchmark sets.
出处
《计算机工程与科学》
CSCD
北大核心
2017年第11期2122-2130,共9页
Computer Engineering & Science
基金
安徽省高校省级重点自然科学研究项目(KJ2013A009)
关键词
差分进化算法
绝对位置
协方差矩阵
旋转不变性
differential evolution algorithm
absolute position
covariance matrix
rotationally invariant