Page 237 - 《软件学报》2026年第2期

软件学报 ISSN 1000-9825, CODEN RUXUEW E-mail: jos@iscas.ac.cn
2026,37(2):716−731 [doi: 10.13328/j.cnki.jos.007440] [CSTR: 32375.14.jos.007440] http://www.jos.org.cn
©中国科学院软件研究所版权所有. Tel: +86-10-62562563

*
面向节点分类的多层异质图神经网络

于朋健, 李 享, 齐建鹏, 于彦伟, 董军宇

(中国海洋大学 信息科学与工程学部, 山东 青岛 266100)
通信作者: 于彦伟, E-mail: yuyanwei@ouc.edu.cn

摘 要: 近年来, 由于异质图卷积网络能够有效学习异质网络语义信息, 逐渐成为网络节点分类的主流算法, 但仍
面临诸多挑战: 现有的大多数工作主要集中在普通异质网络上, 即假设两个节点之间只有一种类型的边, 忽略了多
层异质网络中多类型节点间的多重关系, 以及没有显式地探索不同关系对各类节点表征的影响. 此外, 图神经网络
的过平滑问题也限制了现有模型仅能捕获低阶的局部信息, 几乎无法学习网络的全局相关信息. 为了应对这些挑
战, 提出了一种面向节点分类的多层异质图神经网络 (multiplex heterogeneous graph neural network, MHGNN). 具体
来说, MHGNN 首先学习各类节点在不同关系下的局部初始表征, 再显式地探索不同关系下的表征的重要性以及
有效融合不同关系下各类型节点的表征, 从而捕获多层异质网络中不同交互关系的差异性. 其次, 基于微观经济学
中的替代品和互补品概念, 构造了考虑全局相似性特征的替代品和互补品矩阵, 并通过图神经网络进行信息聚合,
以更好地捕获不同关系下各类节点之间的高阶全局语义信息. 最后, 通过对比学习协调局部和全局两个视图中学
习到的差异性和相似性表征并融合获得最终节点表征. 在 6 个真实数据集上的广泛实验评估证明所提的 MHGNN
在节点分类任务上的各项评估指标都显著优于最新模型.
关键词: 图表示学习; 多层异质网络; 图卷积网络; 节点分类; 对比学习
中图法分类号: TP18

中文引用格式: 于朋健, 李享, 齐建鹏, 于彦伟, 董军宇. 面向节点分类的多层异质图神经网络. 软件学报, 2026, 37(2): 716–731. http://
www.jos.org.cn/1000-9825/7440.htm
英文引用格式: Yu PJ, Li X, Qi JP, Yu YW, Dong JY. Multiplex Heterogeneous Graph Neural Network for Node Classification. Ruan
Jian Xue Bao/Journal of Software, 2026, 37(2): 716–731 (in Chinese). http://www.jos.org.cn/1000-9825/7440.htm

Multiplex Heterogeneous Graph Neural Network for Node Classification
YU Peng-Jian, LI Xiang, QI Jian-Peng, YU Yan-Wei, DONG Jun-Yu
(Faculty of Information Science and Engineering, Ocean University of China, Qingdao 266100, China)
Abstract: In recent years, heterogeneous graph convolutional networks have emerged as a mainstream approach for node classification due
to their ability to effectively capture semantic information in heterogeneous networks. However, several challenges remain. Most existing
studies focus on general heterogeneous networks, where only a single type of edge is assumed between any two nodes. This simplification
overlooks the multiple relationships that exist among multi-type nodes in multiplex heterogeneous networks and fails to explicitly explore
the impact of different relations on the representations of various node types. Moreover, the over-smoothing issue inherent in graph neural
networks limits these models to capturing only low-order local information, making it difficult to learn global correlation patterns in the
network. To address these challenges, this study proposes a multiplex heterogeneous graph neural network (MHGNN) designed for node
classification. The proposed MHGNN first learns local initial representations of each node type under different relational contexts. It then
explicitly models the importance of each relation and effectively integrates the representations of different node types across multiple
relations, thus capturing the relational diversity within multiplex heterogeneous networks. In addition, drawing inspiration from the
microeconomic concepts of substitutes and complements, the study constructs substitute and complement matrices that encode global

* 基金项目: 国家自然科学基金 (62176243, 61773331, 41927805)
收稿时间: 2024-06-25; 修改时间: 2024-10-12; 采用时间: 2025-03-25; jos 在线出版时间: 2025-11-20
CNKI 网络首发时间: 2025-11-21