Gumbel sampling-based differential evolution

摘要：

针对差分进化算法在求解复杂优化问题中面临的搜索多样性不足、易陷入局部最优等问题，提出了一种融合耿贝尔采样机制的差分进化算法,该算法引入了两种新型变异策略:一是基于耿贝尔采样学习的变异策略，通过对高质量个体进行耿贝尔采样，提升个体生成的质量;二是基于耿贝尔采样的精英变异策略，结合耿贝尔扰动机制对精英个体进行局部搜索，增强算法的局部开发能力.两种策略协同作用·能够有效提高种群的多样性和搜索精度，在大量测试集函数和大规模定日镜场应用问题上对所提算法与多种主流差分进化算法变体进行了对比.实验结果表明，所提算法在大多数测试问题上表现优越，兼具较好的全局搜索能力与收敛性能，具有较强的稳定性与适应性.该研究为复杂优化问题的智能求解提供了一种有效的新方法.

To address the issues of insufficient search diversity and premature convergence in differential evolution when solving complex optimization problems, this paper proposes a novel Gumbel sampling based differential evolution. The pro-posed algorithm incorporates two innovative mutation strategies. The first is a Gumbel learning-based mutation strategy, which enhances the quality of generated individuals by applying Gumbel sampling to high- quality solutions. The second is a Gumbel sampling-based elite mutation strategy. which conducts local search around elite individuals using the Gumbel distribution to strengthen local exploitation. The combination of these two strategies effectively improves both population diversity and search precision, Extensive experimnents were conducted on yarious benchmark functions and real-world applications, comparing the proposed algorithm with several state-of- the-art DE variants. Experimental results show that the proposed algorithm outperforms the compared algorithms on most test problems, exhibiting superior global search ability, convergence speed, and ro-bustness, This study provides an effective new approach for solving complex optimization problems with inteligent eyolutionary techniques.

作者: 

张合，王川，黎建宇

Zhang He,Wang Chuan,Li Jianyu

机构地区：

河南师范大学a.教育学部;b.软件学院；南开大学人工智能学院

引用本文：

张合，王川，黎建宇。基于耿贝尔采样的差分进化算法[J].河南师范大学学报(自然科学版).2026.54(2):38-45.(Zhang He,Wang Chuan, Li Jianyu.Gumbel sampling-based differential evolution[J]. Journal of Henan Normal University(Natural Science Edition),2026 ,54(2):38-45.DO1:10.16366/j.cnki.1000-2367.2025.07.11.0001.)

基金：

国家自然科学基金；中央高校基本科研业务费专项资金资助

关键词：

差分进化算法；耿贝尔采样；变异策略；全局优化；进化计算

differential evolution; gumbel sampling; mutation strategy; global optimization; evolutionary computation

分类号：

TP278;TP391

基于耿贝尔采样的差分进化算法.pdf