GNN谱视角下的跨网络节点分类

doi:10.12066/j.issn.1007-2861.2517

摘要/Abstract

摘要： 为了解决小样本问题,旨在分析跨网络节点的频谱信息传递机理,建立基于域适应的跨网络节点分类的频谱方法.在理论上,首先提出了一个中频图卷积核来为图神经网络(graph neural network,GNN)引入中频信息,然后分析了不同频率信息的性质.研究发现,不同于低频和高频信息,中频信息在单网络和跨网络中均保留的是节点的相似特征.这暗示了中频信息比低频和高频信息更适用于跨网络节点分类.在实践上,设计了一种中频域适应图卷积网络(intermediate frequency domainadaptive graph convolutional network,IFDA-GCN)来跨网络分类节点.利用中频图卷积核来提取源网络和目标网络的中频信息,并利用域适应技术来缩小不同网络的特征分布差异.真实数据集上的实验结果表明,IFDA-GCN具有比其他对比方法更优异的性能,并且证实了在跨网络节点分类上,中频信息确实比低频和高频信息有更好的表现.

关键词: 图神经网络, 频率信息, 域适应, 节点分类

Abstract: To solve the small-sample problem, this paper aimed to analyze the mechanism of spectral information transmission across network nodes and established a spectral method for cross-network node classification based on domain adaptation. In theory, this paper proposed an intermediate-frequency graph convolutional kernel to introduce intermediate-frequency information for a graph neural network (GNN) and then analyzed the properties of different frequency information. The results showed that unlike low-frequency and high-frequency information, intermediate-frequency information preserved similar features of nodes both in a single network and across networks. It was demonstrated that intermediate-frequency information was considered more suitable for cross-network node classification than low- frequency and high-frequency information. In practice, an intermediate frequency-domain adaptive graph convolutional network (IFDA-GCN) was proposed for classifying nodes across networks. IFDA-GCN relied on the intermediate-frequency graph convolutional kernel to extract intermediate-frequency information from the source and target networks, while leveraging domain adaptation techniques to mitigate the distributional shift between networks. Experimental results on real-world datasets demonstrated that IFDA-GCN outperformed baselines, and that intermediate-frequency information outperformed low-frequency and high-frequency information in cross-network node classification.

Key words: graph neural network (GNN), frequency information, domain adaptation, node classification

中图分类号:

TP391

朱兆迪, 彭亚新. GNN谱视角下的跨网络节点分类[J]. 上海大学学报(自然科学版), 2026, 32(2): 295-311.

ZHU Zhaodi, PENG Yaxin. Cross-network node classification from the spectral perspective of GNN[J]. Journal of Shanghai University（Natural Science Edition）, 2026, 32(2): 295-311.

参考文献

[1] Pan S J, Yang Q. A survey on transfer learning [J]. IEEE Transactions on Knowledge and Data Engineering, 2010, 22(10): 1345-1359.
[2] Chen Y, Song S, Li S, et al. Domain space transfer extreme learning machine for domain adaptation [J]. IEEE Transactions on Cybernetics, 2019, 49(5): 1909-1922.
[3] Li J, Lu K, Huang Z, et al. Transfer independently together: a generalized framework for domain adaptation [J]. IEEE Transactions on Cybernetics, 2019, 49(6): 2144-2155.
[4] Zhao Y, Zhong Z, Luo Z, et al. Source-free open compound domain adaptation in semantic segmentation [J]. IEEE Transactions on Circuits and Systems for Video Technology, 2022, 32(10): 7019-7032.
[5] Li G, Ji Z, Qu X, et al. Cross-domain object detection for autonomous driving: a stepwise domain adaptative YOLO approach [J]. IEEE Transactions on Intelligent Vehicles, 2022, 7(3): 603-615.
[6] Wu F, Huang Y. Sentiment domain adaptation with multiple sources [C]// Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics. 2016: 301-310.
[7] Wu F, Yuan Z, Huang Y. Collaboratively training sentiment classiflers for multiple domains [J]. IEEE Transactions on Knowledge and Data Engineering, 2017, 29(7): 1370-1383.
[8] Fang M, Yin J, Zhu X. Transfer learning across networks for collective classiflcation [C]// 2013 IEEE 13th International Conference on Data Mining. 2013: 161-170.
[9] Kipf T N, Welling M. Semi-supervised classiflcation with graph convolutional networks [EB/OL]. (2017-02-22) [2023-05-18]. http://arxiv.org/abs/1609.02907.
[10] Bhagat S, Cormode G, Muthukrishnan S. Node classiflcation in social networks [M]. Boston, MA: Springer, 2011: 115-148.
[11] Fout A, Byrd J, Shariat B, et al. Protein interface prediction using graph convolutional networks [J]. Advances in Neural Information Processing Systems, 2017, 30: 6530-6539.
[12] Lecun Y, Bengio Y. Convolutional networks for images, speech, and time series [M]. Cambridge, MA: MIT Press, 1998.
[13] Hochreiter S, Schmidhuber J. Long short-term memory [J]. Neural Computation, 1997, 9(8): 1735-1780.
[14] Wu Z, Pan S, Chen F, et al. A comprehensive survey on graph neural networks [J]. IEEE Transactions on Neural Networks and Learning Systems, 2020, 32(1): 4-24.
[15] Bo D, Wang X, Shi C, et al. Beyond low-frequency information in graph convolutional networks [C]// Proceedings of the AAAI Conference on Artiflcial Intelligence. 2021: 3950-3957.
[16] Nt H, Maehara T. Revisiting graph neural networks: all we have is low-pass fllters [EB/OL]. (2019-05-26) [2023-05-18]. http://arxiv.org/abs/1905.09550.
[17] Zellinger W, Grubinger T, Lughofer E, et al. Central moment discrepancy (CMD) for domain-invariant representation learning [EB/OL]. (2019-05-02) [2023-05-18]. http://arxiv.org/abs/1702.08811.
[18] Cui P, Wang X, Pei J, et al. A survey on network embedding [J]. IEEE Transactions on Knowledge and Data Engineering, 2019, 31(5): 833-852.
[19] Perozzi B, Al-Rfou R, Skiena S. DeepWalk: online learning of social representations [C]// Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 2014: 701-710.
[20] Grover A, Leskovec J. Node2vec: Scalable feature learning for networks [C]// Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 2016: 855-864.
[21] Zhang Z, Yang H, Bu J, et al. ANRL: attributed network representation learning via deep neural networks [C]// International Joint Conference on Artiflcial Intelligence. 2018: 3155-3161.
[22] Huang X, Li J, Hu X. Label informed attributed network embedding [C]// Proceedings of the Tenth ACM International Conference on Web Search and Data Mining. 2017: 731-739.
[23] Hamilton W, Ying Z, Leskovec J. Inductive representation learning on large graphs [J]. Advances in Neural Information Processing Systems, 2017, 30: 1024-1034.
[24] Veličković P, Cucurull G, Casanova A, et al. Graph attention networks [EB/OL]. (2018-02-04) [2023-05-18]. http://arxiv.org/abs/1710.10903.
[25] Gilmer J, Schoenholz S S, Riley P F, et al. Neural message passing for quantum chemistry [C]// International Conference on Machine Learning. 2017: 1263-1272.
[26] Bianchi F M, Grattarola D, Livi L, et al. Graph neural networks with convolutional ARMA fllters [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 44(7): 3496-3507.
[27] He M, Wei Z, Xu H, et al. BernNet: learning arbitrary graph spectral fllters via Bernstein approximation [J]. Advances in Neural Information Processing Systems, 2021, 34: 14239-14251.
[28] Wang X, Zhang M. How powerful are spectral graph neural networks [C]// International Conference on Machine Learning. 2022: 23341-23362.
[29] Shuman D I, Narang S K, Frossard P, et al. The emerging fleld of signal processing on graphs: extending high-dimensional data analysis to networks and other irregular domains [J]. IEEE Signal Processing Magazine, 2013, 30(3): 83-98.
[30] Ding C, Li T, Peng W, et al. Orthogonal nonnegative matrix t-factorizations for clustering [C]// Proceedings of the 12th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 2006: 126-135.
[31] Shen X, Dai Q, Mao S, et al. Network together: node classiflcation via cross-network deep network embedding [J]. IEEE Transactions on Neural Networks and Learning Systems, 2021, 32(5): 1935-1948.
[32] Gretton A, Borgwardt K, Rasch M, et al. A kernel method for the two-sample-problem [J]. Advances in Neural Information Processing Systems, 2006, 19: 513-520.
[33] Dai Q, Wu X M, Xiao J, et al. Graph transfer learning via adversarial domain adaptation with graph convolution [J]. IEEE Transactions on Knowledge and Data Engineering, 2023, 35(5): 4908-4922.
[34] Wu M, Pan S, Zhou C, et al. Unsupervised domain adaptive graph convolutional networks [C]// Proceedings of the Web Conference. 2020: 1457-1467.
[35] Shen J, Qu Y, Zhang W, et al. Wasserstein distance guided representation learning for domain adaptation [C]// Proceedings of the AAAI Conference on Artiflcial Intelligence. 2018: 4058-4065.
[36] Ganin Y, Ustinova E, Ajakan H, et al. Domain-adversarial training of neural networks [J]. Journal of Machine Learning Research, 2016, 17(1): 1-35.
[37] Chung F R, Graham R. Spectral graph theory [M]. Rhode Island, Massachusetts: American Mathematical Society, 1997.
[38] Tang J, Zhang J, Yao L, et al. ArnetMiner: extraction and mining of academic social networks [C]// Proceedings of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 2008: 990-998.