邻域精英集体信息和种群全局信息自适应的多策略差分进化算法  

Adaptive multi-strategy differential evolution algorithm for neighborhood elite collective information and population global information

作　　者：宋晓宇[1] 朱彦霖 赵明[1] Song Xiaoyu;Zhu Yanlin;Zhao Ming(School of Computer Science&Engineering,Shenyang Jianzhu University,Shenyang 110168,China)

机构地区：[1]沈阳建筑大学计算机科学与工程学院,沈阳110168

出　　处：《计算机应用研究》2024年第12期3710-3715,共6页Application Research of Computers

基　　金：辽宁省教育厅重点科研资助项目(lnzd202004);辽宁省科技厅自然科学基金计划资助项目(2023-MS-222)。

摘　　要：为了使差分进化算法(differential evolution,DE)能够更好地利用个体邻域和整个种群的信息,提出了邻域精英信息和种群全局信息自适应的多策略差分进化算法(adaptive multi-strategy differential evolution algorithm for neighborhood elite collective information and population global information,MSDE-NECPG)。首先,充分利用个体邻域中多个精英个体的信息对变异策略进行引导,使搜索向更好的方向移动,提高开发能力。其次,为了让邻域的状态能够随着搜索过程不断地进化,引入邻域更新机制。当邻域最优个体连续多代更新失败,邻域可能陷入局部最优,此时扩大邻域半径,提高探索能力。同时,引入变异策略“DE/current-to-pbest”,这一策略不划分邻域,是基于种群的全局信息。两个策略基于个体的改进率进行多策略的自适应,在局部信息和全局信息之间进行平衡。此外,为了防止参数的错误交互,缩放因子F、交叉率CR根据成功历史积累进行更新,采用分组的参数自适应机制,不断适应搜索过程。最后,为了验证其有效性,在CEC2014的30个基准函数上,与5种迄今为止比较先进的差分进化算法进行比较,实验结果表明,所提算法的精度、稳定性和收敛速度比得上这5种先进的算法。This paper proposed a multi-strategy differential evolution algorithm(MSDE-NECPG)to better utilize individual neighborhood and global population information in the differential evolution(DE)algorithm.Firstly,it fully utilized the information of multiple elite individuals in the individual neighborhood to guide the mutation strategy,moving the search towards better directions and enhancing the exploitation capability.Secondly,it introduced an update mechanism for the neighborhood to ensure its state evolved continuously throughout the search process.When the optimal individual in the neighborhood fails to update consecutively for multiple generations,the neighborhood may fall into a local optimum.At this point,this paper expanded the neighborhood radius to increase exploration capability.It introduced the mutation strategy“DE/current-to-pbest”,which was based on the global information of the population.These two strategies adaptively adjusted based on the improvement rate of individuals,balancing between local and global information.Furthermore,to prevent parameter error interaction,the scaling factor F and crossover rate CR were updated based on successful historical accumulations,employing a grouped parameter adaptive mechanism to continuously adapt to the search process.Finally,to validate its effectiveness,experiments were conducted on 30 benchmark functions from CEC2014,comparing it with five state-of-the-art differential evolution algorithms.The experimental results demonstrate that the proposed algorithm is comparable to these five advanced algorithms in terms of accuracy,stability and convergence speed.

关 键 词：差分进化 邻域精英信息 多策略自适应 参数自适应 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]