《计算机应用》唯一官方网站 ›› 2025, Vol. 45 ›› Issue (1): 15-23.DOI: 10.11772/j.issn.1001-9081.2023121847

• 人工智能 • 上一篇    下一篇

基于有向超图自适应卷积的链接预测模型

赵文博1,2, 马紫彤1,2, 杨哲1,2()   

  1. 1.苏州大学 计算机科学与技术学院,江苏 苏州 215008
    2.江苏省计算机信息处理技术重点实验室(苏州大学),江苏 苏州 215006
  • 收稿日期:2024-01-04 修回日期:2024-03-12 接受日期:2024-03-15 发布日期:2024-03-28 出版日期:2025-01-10
  • 通讯作者: 杨哲
  • 作者简介:赵文博(1998—),男,安徽淮北人,硕士研究生,CCF会员,主要研究方向:图机器学习、推荐系统;
    马紫彤(1999—),女,江西赣州人,硕士研究生,CCF会员,主要研究方向:图表示学习、深度学习、推荐系统;
  • 基金资助:
    国家自然科学基金资助项目(62272332);教育部产学合作协同育人项目(220606363154256)

Link prediction model based on directed hypergraph adaptive convolution

Wenbo ZHAO1,2, Zitong MA1,2, Zhe YANG1,2()   

  1. 1.School of Computer Science and Technology,Soochow University,Suzhou Jiangsu 215008,China
    2.Jiangsu Provincial Key Laboratory for Computer Information Processing Technology (Soochow University),Suzhou Jiangsu 215006,China
  • Received:2024-01-04 Revised:2024-03-12 Accepted:2024-03-15 Online:2024-03-28 Published:2025-01-10
  • Contact: Zhe YANG
  • About author:ZHAO Wenbo, born in 1998, M. S. candidate. His research interests include graph machine learning, recommender systems.
    MA Zitong, born in 1999, M. S. candidate. Her research interests include graph representation learning, deep learning, recommender systems.
  • Supported by:
    National Natural Science Foundation of China(62272332);Ministry of Education’s Collaborative Education Program on Industry and Education(220606363154256)

摘要:

图神经网络(GNN)为链接预测提供了多样化的解决方案,但由于普通图的结构限制,目前的相关模型在充分利用顶点间的高阶及不对称信息方面存在明显的不足。针对以上问题,提出一种基于有向超图自适应卷积的链接预测模型。首先,使用有向超图结构更充分地表示顶点间的高阶和方向信息,兼具超图和有向图的优势;其次,有向超图自适应卷积采用自适应信息传播方式替代传统有向超图中的定向信息传播方式,从而解决了有向超边尾部顶点不能有效更新嵌入的问题,同时解决多层卷积导致的顶点过度平滑问题。在Citeseer数据集上基于显式顶点特征的实验结果显示,在链接预测任务上,相较于有向超图神经网络(DHNN)模型,所提模型的ROC (Receiver Operating Characteristic)曲线下面积(AUC)指标提升了2.23个百分点,平均精度(AP)提升了1.31个百分点。因此,所提模型可以充分表达顶点间的关系,并有效提高链接预测任务的性能。

关键词: 图神经网络, 有向超图, 链接预测, 超图卷积, 表示学习, 自适应卷积

Abstract:

Although diverse solutions for link prediction have been provided by Graph Neural Networks (GNN), the recent models have significant shortcomings in fully utilizing high-order and asymmetric information between vertices caused by the structural constraints of ordinary graphs. To address the above issues, a link prediction model based on directed hypergraph adaptive convolution was proposed. Firstly, the directed hypergraph structure was employed to represent high-order and directional information between vertices more sufficiently, possessing advantages of both hypergraphs and directed graphs. Secondly, an adaptive information propagation method was adopted by directed hypergraph adaptive convolution to replace the directional information propagation method in traditional directed hypergraphs, thereby solving the problem of ineffective updating of embeddings for tail vertices of directed hyperedges, and solving the problem of excessive smoothing of vertices caused by multi-layer convolution. Experimental results based on explicit vertex features on Citeseer dataset show that the proposed model achieves a 2.23 percentage points increase in the Area Under the ROC (Receiver Operating Characteristic) Curve (AUC) and a 1.31 percentage points increase in Average Precision (AP) compared to the Directed Hypergraph Neural Network (DHNN) model in link prediction task. Therefore, it can be concluded that this model expresses the relationships between vertices adequately and improves the accuracy of link prediction task effectively.

Key words: Graph Neural Network (GNN), directed hypergraph, link prediction, hypergraph convolution, representation learning, adaptive convolution

中图分类号: