Scholarlink

<p>The exponential distribution optimizer (EDO) is a swarm intelligence algorithm based on an exponential probability distribution model. While it excels in global exploration, the original EDO struggles with local rigorous exploitation, potentially causing insufficient population diversity, susceptibility to local optima, and slow convergence. To this end, in this paper, an enhanced variant of EDO called TEDO is proposed to address these shortcomings by introducing several strategies. Firstly, Latin hypercube sampling (LHS) enhances the diversity of the initial population. Secondly, the adaptive <i>t</i>-distribution strategy and the mean adaptive trap avoidance strategy (MTAO) help to jump out of local optima. Thirdly, <i>S</i>-type escape balances the two individual updating methods in the exploitation phase. Finally, the elite-oriented guided solution improves the convergence behavior of the algorithm. This study evaluates the performance of the TEDO algorithm using the CEC2017 and CEC2022 benchmark test suites, along with 19 mechanical engineering problems from the CEC2020 real-world optimization suite. The effectiveness of TEDO is further assessed through comparisons with ten other algorithms, including EDO and its variants. Experimental results demonstrate improved convergence speed and accuracy of TEDO. The Wilcoxon signed-rank test results on CEC2017 (10, 30, 50, 100<i>D</i>) and CEC2022 (10, 20<i>D</i>) show performance improvements, with results of 256/29/5, 275/11/4, 279/9/2, 280/6/4, 107/8/5, and 110/10/0, respectively, indicating the potential effectiveness of TEDO.</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

A multi-strategy improved exponential distribution optimizer for numerical optimizations and engineering problems

  • 期刊论文

  • 发表日期：

    2025年6月7日

查看全文
  • Tengfei Yang,
  • Hao Liu

摘要

The exponential distribution optimizer (EDO) is a swarm intelligence algorithm based on an exponential probability distribution model. While it excels in global exploration, the original EDO struggles with local rigorous exploitation, potentially causing insufficient population diversity, susceptibility to local optima, and slow convergence. To this end, in this paper, an enhanced variant of EDO called TEDO is proposed to address these shortcomings by introducing several strategies. Firstly, Latin hypercube sampling (LHS) enhances the diversity of the initial population. Secondly, the adaptive t-distribution strategy and the mean adaptive trap avoidance strategy (MTAO) help to jump out of local optima. Thirdly, S-type escape balances the two individual updating methods in the exploitation phase. Finally, the elite-oriented guided solution improves the convergence behavior of the algorithm. This study evaluates the performance of the TEDO algorithm using the CEC2017 and CEC2022 benchmark test suites, along with 19 mechanical engineering problems from the CEC2020 real-world optimization suite. The effectiveness of TEDO is further assessed through comparisons with ten other algorithms, including EDO and its variants. Experimental results demonstrate improved convergence speed and accuracy of TEDO. The Wilcoxon signed-rank test results on CEC2017 (10, 30, 50, 100D) and CEC2022 (10, 20D) show performance improvements, with results of 256/29/5, 275/11/4, 279/9/2, 280/6/4, 107/8/5, and 110/10/0, respectively, indicating the potential effectiveness of TEDO.