Scholarlink

<p>Particle swarm optimization (PSO) algorithm is one of the most popular heuristic stochastic search algorithms, and has been successfully applied in various scientific and engineering fields. However, when dealing with complex problems, it still suffers in low search effectiveness. To address this drawback, a novel PSO variant, named enhanced particle swarm algorithm with diversity-based adaptive predicted learning strategy, is proposed in this paper. Particularly, aiming at strengthening the search quality of algorithm, two predicted-based velocity update strategies are first presented by making full use of the neighborhood information or population information to predict the promising regions and construct the search direction for particles. Moreover, in order to well maintain the population diversity, a diversity-based adaptive mixed learning strategy is then developed, where the dimensional diversity of population is measured and adopted to choose the proper method among the proposed two ones above to update the corresponding dimension of particle. Additionally, to further integrate the benefits of the traditional PSO algorithm and the above mixed strategy, a learning strategy selection mechanism is further devised by using the update of particle to select the suitable approach among them for it. Unlike the existing PSO algorithms, the new algorithm introduces the predicted information to guide the search, considers the diversity of dimension to adjust the search ability, and employs the update of particle to choose the proper update approach during the search process. Thereby, it is capable of more availably enhancing the search efficiency of algorithm, and balancing the exploration and exploitation. Finally, the performance of our new algorithm is demonstrated by comparing with 10 typical and up-to-date algorithms on the benchmark functions from IEEE CEC2017 test suite. Compared to these opponents, the proposed algorithm gets significantly better results on 20 out of 30 cases based on the multiproblem Wilcoxon signed-rank test at a significant level of 0.05. Thus, the proposed algorithm has a competitive performance.</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

Enhanced particle swarm algorithm with diversity-based adaptive predicted learning strategy for numerical optimization

  • 期刊论文

  • 发表日期：

    2025年8月30日

查看全文
  • Xuan Liu,
  • Mengnan Tian,
  • Xinduan Wang

摘要

Particle swarm optimization (PSO) algorithm is one of the most popular heuristic stochastic search algorithms, and has been successfully applied in various scientific and engineering fields. However, when dealing with complex problems, it still suffers in low search effectiveness. To address this drawback, a novel PSO variant, named enhanced particle swarm algorithm with diversity-based adaptive predicted learning strategy, is proposed in this paper. Particularly, aiming at strengthening the search quality of algorithm, two predicted-based velocity update strategies are first presented by making full use of the neighborhood information or population information to predict the promising regions and construct the search direction for particles. Moreover, in order to well maintain the population diversity, a diversity-based adaptive mixed learning strategy is then developed, where the dimensional diversity of population is measured and adopted to choose the proper method among the proposed two ones above to update the corresponding dimension of particle. Additionally, to further integrate the benefits of the traditional PSO algorithm and the above mixed strategy, a learning strategy selection mechanism is further devised by using the update of particle to select the suitable approach among them for it. Unlike the existing PSO algorithms, the new algorithm introduces the predicted information to guide the search, considers the diversity of dimension to adjust the search ability, and employs the update of particle to choose the proper update approach during the search process. Thereby, it is capable of more availably enhancing the search efficiency of algorithm, and balancing the exploration and exploitation. Finally, the performance of our new algorithm is demonstrated by comparing with 10 typical and up-to-date algorithms on the benchmark functions from IEEE CEC2017 test suite. Compared to these opponents, the proposed algorithm gets significantly better results on 20 out of 30 cases based on the multiproblem Wilcoxon signed-rank test at a significant level of 0.05. Thus, the proposed algorithm has a competitive performance.