摘要 异质图神经网络在挖掘复杂图数据任务中性能较优,但现有方法主要采用有监督学习范式,高度依赖节点标注信息,对原始图结构数据中的噪声链接较敏感,限制其在标注稀缺场景下的应用.针对上述问题,文中提出基于对比学习和结构更新机制的异质图结构学习方法(Heterogeneous Graph Structure Learning Based on Contrastive Learning and Structure Update Mechanism, HGSL-CL).首先,从原始数据中生成学习目标作为锚视图,结合类型感知特征映射与加权多视角相似度计算,生成学习者视图.然后,通过结构更新机制迭代优化锚视图,使用语义级注意力得到两个视角下的节点表示.最后,使用多层感知机将节点表示投影至同一维度空间,通过跨视角协同对比损失函数实现图结构优化,并引入融合节点拓扑相似度与属性相似度的正样本筛选策略,增强对比学习的判别能力.在3个数据集上的实验表明,HGSL-CL在节点分类、聚类等任务中性能较优,学习的图结构可泛化至半监督场景,取得比原始基线模型更优的性能,由此证实图结构学习的有效性.HGSL-CL源代码: https://github.com/desslie047/HGSL-CL.
Abstract：Heterogeneous graph neural networks hold significant advantages in complex graph data mining tasks. However, existing methods typically follow a supervised learning paradigm. Therefore, they are highly dependent on node labeling information and sensitive to noisy links in the original graph structure. As a result, their applications in labeling-scarce scenarios are limited. To address these issues, a method for heterogeneous graph structure learning based on contrastive learning and structure update mechanism(HGSL-CL) is proposed. The learning target is first generated as the anchor view from the original data. The type-aware feature mapping and weighted multi-view similarity computation are combined to generate the learner view. Subsequently, the anchor view is iteratively optimized through the structure update mechanism, and the node representations in two views are obtained using semantic-level attention. Finally, node representations from both views are projected into a shared latent space via a multi-layer perceptron. The graph structure optimization is achieved by the cross-view synergistic contrastive loss function, and a positive sample filtering strategy fusing node topological similarity and attribute similarity is introduced to enhance the discriminative ability of contrastive learning. Experiments on three datasets show that HGSL-CL outperforms other baseline models in node classification and clustering tasks. Moreover, the learned graph structure can be generalized to semi-supervised scenarios, and HGSL-CL achieves better performance than the original baseline models. The results demonstrate the effectiveness of graph structure learning. The source code of HGSL-CL is available at https://github.com/desslie047/HGSL-CL.
[1] 章淯淞,夏鸿斌,刘渊.结合自对比图神经网络与双预测器的会话推荐模型.模式识别与人工智能, 2024, 37(3): 242-252. (ZHANG Y S, XIA H B, LIU Y.Session-Based Recommendation Model with Self Contrastive Graph Neural Network and Dual Predictor. Pattern Recognition and Artificial Intelligence, 2024, 37(3):242-252.) [2] 张玉朋,李香菊,李超,等.基于Transformer与异质图神经网络的新闻推荐模型.模式识别与人工智能, 2022, 35(9): 839-848. (ZHANG Y P, LI X J, LI C, et al. News Recommendation Model Based on Transformer and Heterogeneous Graph Neural Network. Pattern Recognition and Artificial Intelligence, 2022, 35(9): 839-848.) [3] 邬锦琛,杨兴耀,于炯,等.双通道异构图神经网络序列推荐算法.计算机科学与探索, 2023, 17(6): 1473-1486. (WU J C, YANG X Y, YU J, et al. Sequence Recommendation with Dual Channel Heterogeneous Graph Neural Network. Journal of Frontiers of Computer Science and Technology, 2023, 17(6): 1473-1486.) [4] ZHANG C X, SONG D J, HUANG C, et al. Heterogeneous Graph Neural Network // Proc of the 25th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York, USA: ACM, 2019: 793-803. [5] WANG X, JI H Y, SHI C, et al. Heterogeneous Graph Attention Network // Proc of the World Wide Web Conference. New York, USA: ACM, 2019: 2022-2032. [6] HU Z N, DONG Y X, WANG K S, et al. Heterogeneous Graph Transformer // Proc of the Web Conference. New York, USA: ACM, 2020: 2704-2710. [7] FU X Y, ZHANG J N, MENG Z Q, et al. MAGNN: Metapath Aggre-gated Graph Neural Network for Heterogeneous Graph Embedding // Proc of the Web Conference. New York, USA: ACM, 2020: 2331-2341. [8] LIU Y X, ZHENG Y, ZHANG D K, et al. Towards Unsupervised Deep Graph Structure Learning // Proc of the ACM Web Confe-rence. New York, USA: ACM, 2022: 1392-1403. [9] 李雅琪,王杰,王锋,等.基于层次对比学习的半监督节点分类算法.模式识别与人工智能, 2023, 36(8): 712-720. (LI Y Q, WANG J, WANG F, et al. Semi-supervised Node Classification Algorithm Based on Hierarchical Contrastive Learning. Pa-ttern Recognition and Artificial Intelligence, 2023, 36(8): 712-720.) [10] HU Y F, ZHANG Y.Graph Contrastive Learning with Local and Global Mutual Information Maximization // Proc of the 8th International Conference on Information Technology(IoT and Smart City). New York, USA: ACM, 2021: 74-78. [11] JIN W, MA Y, LIU X R, et al. Graph Structure Learning for Robust Graph Neural Networks // Proc of the 26th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York, USA: ACM, 2020: 66-74. [12] FRANCESCHI L, NIEPERT M, PONTIL M, et al. Learning Discrete Structures for Graph Neural Networks. Proceedings of Machine Learning Research, 2019, 97: 1972-1982. [13] JIANG B, ZHANG Z Y, LIN D D, et al. Semi-supervised Lear-ning with Graph Learning-Convolutional Networks // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2019: 11305-11312. [14] CHEN Y, WU L F, ZAKI M J.Deep Iterative and Adaptive Lear-ning for Graph Neural Networks[C/OL]. [2025-02-24].https://arxiv.org/pdf/1912.07832. [15] SHI C, YU P S.Heterogeneous Information Network Analysis and Applications. Berlin, Germany: Springer, 2017. [16] DONG Y X, CHAWLA N V, SWAMI A. metapath2vec: Scalable Representation Learning for Heterogeneous Networks // Proc of the 23rd ACM SIGKDD International Conference on Knowledge Disco-very and Data Mining. New York, USA: ACM, 2017: 135-144. [17] REN Y X, LIU B, HUANG C, et al. Heterogeneous Deep Graph Infomax[C/OL].[2025-02-24]. https://arxiv.org/pdf/1911.08538. [18] PARK C, KIM D, HAN J W, et al. Unsupervised Attributed Multiplex Network Embedding. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(4): 5371-5378. [19] WANG X, LIU N, HAN H, et al. Self-Supervised Heterogeneous Graph Neural Network with Co-contrastive Learning // Proc of the 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York, USA: ACM, 2021: 1726-1736. [20] ZHAO J N, WANG X, SHI C, et al. Network Schema Preserving Heterogeneous Information Network Embedding // Proc of the 29th International Joint Conference on Artificial Intelligence. San Francisco, USA: IJCAI, 2020: 1366-1372. [21] WANG Z H, LI Q, YU D H, et al. Heterogeneous Graph Contrastive Multi-view Learning // Proc of the SIAM International Confe-rence on Data Mining. Philadelphia, USA: SIAM, 2023: 136-144. [22] YU J X, GE Q Q, LI X, et al. Heterogeneous Graph Contrastive Learning with Meta-Path Contexts and Adaptively Weighted Negative Samples. IEEE Transactions on Knowledge and Data Enginee-ring, 2024, 36(10): 5181-5193. [23] YANG Y M, GUAN Z Y, LI J X, et al. Interpretable and Efficient Heterogeneous Graph Convolutional Network. IEEE Transactions on Knowledge and Data Engineering, 2023, 35(2): 1637-1650.
2025, Vol. 38 
