Scholarlink

<p>Bipartite networks containing two types of nodes are commonly encountered in practice. To analyze bipartite networks, network embedding models are popularly used. With the increase in data availability, nodes in networks are often observed with nodal attributes, which provide fertile information for understanding the network structure. However, existing network embedding models for dynamic bipartite networks often ignore nodal variables. To address this problem, we propose a latent space model for bipartite networks by incorporating information from both the covariates and the network structure. To reflect the evolution pattern of the network structure, we introduce two parameters representing the persistence effect. To estimate the model, we propose a computationally efficient algorithm using the projected gradient descent method. The theoretical properties are also established and validated through comprehensive simulation studies. Last, we analyze the dynamic bipartite network for the Chinese interlocking directorates from 2010 to 2020 using our proposed model.</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

Network embedding for bipartite networks with applications in interlocking directorates in Chinese companies

  • 期刊论文

  • 发表日期：

    2025年4月27日

查看全文
  • Yan Zhang,
  • Rui Pan,
  • Feifei Wang,
  • Kuangnan Fang,
  • Hansheng Wang

摘要

Bipartite networks containing two types of nodes are commonly encountered in practice. To analyze bipartite networks, network embedding models are popularly used. With the increase in data availability, nodes in networks are often observed with nodal attributes, which provide fertile information for understanding the network structure. However, existing network embedding models for dynamic bipartite networks often ignore nodal variables. To address this problem, we propose a latent space model for bipartite networks by incorporating information from both the covariates and the network structure. To reflect the evolution pattern of the network structure, we introduce two parameters representing the persistence effect. To estimate the model, we propose a computationally efficient algorithm using the projected gradient descent method. The theoretical properties are also established and validated through comprehensive simulation studies. Last, we analyze the dynamic bipartite network for the Chinese interlocking directorates from 2010 to 2020 using our proposed model.