Scholarlink

<p>In this paper, we propose an improved spatial information sampling (SIS) algorithm based on an adaptive distributed structure. The original SIS divides the population into internal and external parts, which perform exploitation and exploration, respectively. Depending on the position of the best individual, SIS activates the corresponding search operator. Although SIS achieves fast convergence and strong optimization ability, it often fails to reach high-quality solutions and may fall into local optima when facing complex problems. To address these limitations, we introduce a new adaptive distributed architecture with multilayer information interaction and integrate it into SIS. This enhancement leads to the spatial information sampling algorithm with adaptive distributed population structure (DMSIS). We conduct extensive experiments comparing DMSIS with the original SIS and seven other advanced algorithms. Our evaluation includes 29 benchmark functions from IEEE CEC2017, 22 real-world optimization tasks from IEEE CEC2011, and 12 benchmark functions from IEEE CEC2022, covering a total of 63 optimization problems. We also conducted ablation studies on the adaptive distributed population structure. The results confirm that although both the adaptive distributed structure and the multilayer interaction mechanism are effective on their own, their combination yields the best overall performance. Statistical tests based on the Wilcoxon rank-sum method show that DMSIS achieves better global search performance and avoids local optima more effectively than competing algorithms.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Spatial information sampling algorithm with adaptive distributed population structure for optimization

  • 期刊论文

  • 发表日期：

    2026年2月7日

查看全文
  • Qiong Fu,
  • Haotian Li,
  • Houtian He,
  • Zhenyu Lei,
  • Shangce Gao

摘要

In this paper, we propose an improved spatial information sampling (SIS) algorithm based on an adaptive distributed structure. The original SIS divides the population into internal and external parts, which perform exploitation and exploration, respectively. Depending on the position of the best individual, SIS activates the corresponding search operator. Although SIS achieves fast convergence and strong optimization ability, it often fails to reach high-quality solutions and may fall into local optima when facing complex problems. To address these limitations, we introduce a new adaptive distributed architecture with multilayer information interaction and integrate it into SIS. This enhancement leads to the spatial information sampling algorithm with adaptive distributed population structure (DMSIS). We conduct extensive experiments comparing DMSIS with the original SIS and seven other advanced algorithms. Our evaluation includes 29 benchmark functions from IEEE CEC2017, 22 real-world optimization tasks from IEEE CEC2011, and 12 benchmark functions from IEEE CEC2022, covering a total of 63 optimization problems. We also conducted ablation studies on the adaptive distributed population structure. The results confirm that although both the adaptive distributed structure and the multilayer interaction mechanism are effective on their own, their combination yields the best overall performance. Statistical tests based on the Wilcoxon rank-sum method show that DMSIS achieves better global search performance and avoids local optima more effectively than competing algorithms.