《计算机应用》唯一官方网站 ›› 2026, Vol. 46 ›› Issue (10): 3083-3092.DOI: 10.11772/j.issn.1001-9081.2025081065

• 人工智能 • 上一篇    

基于多头关系转化的异构图神经网络

黄华娟, 方永康, 韦修喜()   

  1. 广西民族大学 人工智能学院,南宁 530006
  • 收稿日期:2025-09-17 修回日期:2025-12-07 接受日期:2025-12-12 发布日期:2026-02-12 出版日期:2026-10-10
  • 通讯作者: 韦修喜
  • 作者简介:黄华娟(1984—),女,广西崇左人,教授,博士,CCF会员,主要研究方向:机器学习、数据挖掘
    方永康(2000—),男,湖北荆州人,硕士研究生,主要研究方向:机器学习
  • 基金资助:
    国家自然科学基金资助项目(62266007);广西自然科学基金资助项目(2021GXNSFAA220068)

Heterogeneous graph neural network based on multi-head relation transformation

Huajuan HUANG, Yongkang FANG, Xiuxi WEI()   

  1. School of Artificial Intelligence,Guangxi Minzu University,Nanning Guangxi 530006,China
  • Received:2025-09-17 Revised:2025-12-07 Accepted:2025-12-12 Online:2026-02-12 Published:2026-10-10
  • Contact: Xiuxi WEI
  • About author:HUANG Huajuan, born in 1984, Ph. D., professor. Her research interests include machine learning, data mining.
    FANG Yongkang, born in 2000, M. S. candidate. His research interests include machine learning.
  • Supported by:
    National Natural Science Foundation of China(62266007);Guangxi Natural Science Foundation(2021GXNSFAA220068)

摘要:

针对异构图神经网络(HGNN)在多关系语义建模中依赖人工构造元路径、难以充分捕捉复杂关系语义的问题,提出一种基于多头关系转化的异构图神经网络(MHRTHGN)方法。该方法为不同关系构建独立的关系变换矩阵,引入多头注意力机制精细刻画多关系特征,并设计关系特定多头注意力机制在节点级与关系级两层面融合多关系信息,同时结合前馈网络(FFN)进一步提升表示的语义表达能力。实验结果表明,在ACM、IMDB与DBLP数据集上,与HAN(Heterogeneous graph Attention Network)、图卷积网络(GCN)等代表性模型相比,所提方法在节点分类与节点聚类任务中的准确率平均提升3~5个百分点,验证了模型在多关系异构图学习中的有效性和性能优势。可见,MHRTHGN能在无须元路径先验的条件下有效建模多关系信息,为异构图表示学习提供了一种泛化性更强、应用潜力更大的技术方案。

关键词: 异构图, 图神经网络, 图表示学习, 多头注意力机制, 关系感知

Abstract:

Concerning the problem that the existing Heterogeneous Graph Neural Networks (HGNNs) rely on manually constructed meta-paths and fail to fully capture complex relation semantics, a Multi-Head Relation-Transformed Heterogeneous Graph neural Network (MHRTHGN) method was proposed. In this method, independent relation transformation matrices were constructed for each type of relation, a multi-head attention mechanism was introduced to characterize multi-relation features precisely, and a relation-specific multi-head attention mechanism was designed to fuse multi-relation information at both node-level and relation-level. At the same time, a Feed-Forward Network (FFN) was combined to further enhance semantic expressiveness capability of the model. Experimental results show that on the ACM, IMDB, and DBLP datasets, the proposed method improves the accuracies of node classification and clustering tasks by 3-5 percentage points on average compared with representative baseline models such as HAN (Heterogeneous graph Attention Network) and Graph Convolutional Network (GCN), validating its effectiveness and performance superiority in multi-relation heterogeneous graph learning. It can be seen that MHRTHGN method can model multi-relation information without relying on meta-path prior, providing a more generalizable solution with higher application potential for heterogeneous graph representation learning.

Key words: Heterogeneous Graph (HG), Graph Neural Network (GNN), graph representation learning, multi-head attention mechanism, relation-awareness

中图分类号: