Scholarlink

<p>As a pivotal problem in social network analysis, the influence maximization problem aims to find a set of key users from target social networks to maximize the spread of information. Existing evolutionary optimizers mainly conduct one specific local search strategy merely within a single-neighborhood to improve the solution quality of the current generation in addressing the problem. However, more promising alternatives tend to be missed in the semi-local or even the global regions, especially in heterogeneous network structures. To address such challenges, a variable neighborhood search-assisted discrete particle swarm optimization (VNSDPSO) is proposed in this paper. The meta-heuristic optimizer maintains an evolutionary population and employs a random partitioning mechanism to divide the population into two subpopulations during the evolution: an excellent subpopulation and a common subpopulation. A general local search strategy is adopted to refine the solutions of the excellent subpopulation, while a lightweight variant of variable neighborhood search strategy is implemented on the common subpopulation to overcome the limitations of the traditional hidebound exploration. Such strategy adjusts the searching neighborhood dynamically to enhance the quality of solution and speed up the convergence. Empirical evaluations on three real-world and three synthetic networks demonstrate the superiority of the proposed VNSDPSO, showing an average 6% enhancement in influence spread against leading baselines.</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

Reinforce the particle swarm optimization with variable neighborhood search for the influence maximization problem in social networks

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月28日

查看全文
  • Jianxin Tang,
  • Xinyue Li,
  • Huiyi Wei,
  • Juan Pang,
  • Chen Wu

摘要

As a pivotal problem in social network analysis, the influence maximization problem aims to find a set of key users from target social networks to maximize the spread of information. Existing evolutionary optimizers mainly conduct one specific local search strategy merely within a single-neighborhood to improve the solution quality of the current generation in addressing the problem. However, more promising alternatives tend to be missed in the semi-local or even the global regions, especially in heterogeneous network structures. To address such challenges, a variable neighborhood search-assisted discrete particle swarm optimization (VNSDPSO) is proposed in this paper. The meta-heuristic optimizer maintains an evolutionary population and employs a random partitioning mechanism to divide the population into two subpopulations during the evolution: an excellent subpopulation and a common subpopulation. A general local search strategy is adopted to refine the solutions of the excellent subpopulation, while a lightweight variant of variable neighborhood search strategy is implemented on the common subpopulation to overcome the limitations of the traditional hidebound exploration. Such strategy adjusts the searching neighborhood dynamically to enhance the quality of solution and speed up the convergence. Empirical evaluations on three real-world and three synthetic networks demonstrate the superiority of the proposed VNSDPSO, showing an average 6% enhancement in influence spread against leading baselines.