Scholarlink

<p>Although existing research has confirmed the importance of higher-order structures in identifying key nodes within networks, the challenge remains on how to effectively integrate different types of higher-order information to precisely locate nodes that may be inconspicuous in lower-order structures but play a crucial role in higher-order interactions. To address this challenge, this paper proposes a general Higher-order Graph Neural Network representation learning framework (HoGNN) that can flexibly adapt to various types of higher-order relationships. Based on a robust theoretical framework, we develop a network dismantling model, SPR(Structural and Processual Role-aware Network Dismantling), which integrates multi-dimensional higher-order relations from both macro and micro perspectives. Empirical analysis demonstrated that the proposed model exhibits superior dismantling efficiency on both real-world and synthetic networks, using the minimal number of target node removals to collapse the network. Moreover, we show that SPR is more resilient to interference and accurately identifies key nodes in networks with multi-dimensional complex structures.</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

Dismantling complex networks based on higher-order graph neural network

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月27日

查看全文
  • Wennan Zhou,
  • Suoyi Tan,
  • Yang Fang,
  • Xin Lü,
  • Xiang Zhao

摘要

Although existing research has confirmed the importance of higher-order structures in identifying key nodes within networks, the challenge remains on how to effectively integrate different types of higher-order information to precisely locate nodes that may be inconspicuous in lower-order structures but play a crucial role in higher-order interactions. To address this challenge, this paper proposes a general Higher-order Graph Neural Network representation learning framework (HoGNN) that can flexibly adapt to various types of higher-order relationships. Based on a robust theoretical framework, we develop a network dismantling model, SPR(Structural and Processual Role-aware Network Dismantling), which integrates multi-dimensional higher-order relations from both macro and micro perspectives. Empirical analysis demonstrated that the proposed model exhibits superior dismantling efficiency on both real-world and synthetic networks, using the minimal number of target node removals to collapse the network. Moreover, we show that SPR is more resilient to interference and accurately identifies key nodes in networks with multi-dimensional complex structures.