Training Robust Graph Completion Networks With Extremely Weak Supervision on Graphs With Incomplete Features and Structure

Chengxiang Lei; Sichao Fu; Qinmu Peng; Yiyang Zhang; Bin Zou; Xiao-Yuan Jing; Xinge You

doi:10.1109/JAS.2025.125906

Volume 13 Issue 7

Jul. 2026

IEEE/CAA Journal of Automatica Sinica

JCR Impact Factor: 18.3, Top 1 (SCI Q1)

CiteScore: 28.2, Top 1% (Q1)
Google Scholar h5-index: 95， TOP 5

Turn off MathJax

Article Contents

Article Navigation > IEEE/CAA Journal of Automatica Sinica > 2026 > 13(7): 1707-1720

C. Lei, S. Fu, Q. Peng, Y. Zhang, B. Zou, X.-Y. Jing, and X. You, “Training robust graph completion networks with extremely weak supervision on graphs with incomplete features and structure,” IEEE/CAA J. Autom. Sinica, vol. 13, no. 7, pp. 1707–1720, Jul. 2026. doi: 10.1109/JAS.2025.125906

Citation:

C. Lei, S. Fu, Q. Peng, Y. Zhang, B. Zou, X.-Y. Jing, and X. You, “Training robust graph completion networks with extremely weak supervision on graphs with incomplete features and structure,” IEEE/CAA J. Autom. Sinica, vol. 13, no. 7, pp. 1707–1720, Jul. 2026. doi: 10.1109/JAS.2025.125906

Citation:

C. Lei, S. Fu, Q. Peng, Y. Zhang, B. Zou, X.-Y. Jing, and X. You, “Training robust graph completion networks with extremely weak supervision on graphs with incomplete features and structure,” IEEE/CAA J. Autom. Sinica, vol. 13, no. 7, pp. 1707–1720, Jul. 2026. doi: 10.1109/JAS.2025.125906

PDF( 2217 KB)

Training Robust Graph Completion Networks With Extremely Weak Supervision on Graphs With Incomplete Features and Structure

doi: 10.1109/JAS.2025.125906

Funds: This work was supported in part by the General Program of the National Natural Science Foundation of China (62575116), the Open Project of the Text Computing and Cognitive Intelligence Ministry of Education Engineering Research Center (TCCI250208), and the Fundamental Research Funds for the Central Universities (2024JYCXJJ062)

More Information

Author Bio:
Chengxiang Lei received the M.S. degree in information and communication engineering from the School of Electronic Information and Communications, Huazhong University of Science and Technology in 2024. He is currently a Researcher with the Electric Power Research Institute of Guangdong Power Grid Co., Ltd. His current research interests include machine learning and applications

Sichao Fu is currently a Ph.D. degree candidate in information and communication engineering at the School of Electronic Information and Communications, Huazhong University of Science and Technology. His research interests include pattern recognition and deep manifold learning

Qinmu Peng received the Ph.D. degree in computer science from the Department of Computer Science, Hong Kong Baptist University China, in 2015. He is currently an Associate Professor with the School of Electronic Information and Communications, Huazhong University of Science and Technology. His current research interests include medical image processing, pattern recognition, machine learning, and computer vision

Yiyang Zhang is currently a Ph.D. degree candidate in information and communication engineering at the School of Electronic Information and Communications, Huazhong University of Science and Technology. His research interests include model security research and evaluation

Bin Zou received the M.S. degree in applied mathematics and the Ph.D. degree in pure mathematics from Hubei University in 2004 and 2007, respectively. He is currently a Professor with the Key Laboratory of Applied Mathematics and the Faculty of Mathematics and Statistics, Hubei University. His current research interests include statistical learning theory, machine learning and pattern recognition

Xiao-Yuan Jing received the Ph.D. degree in pattern recognition and intelligent systems from Nanjing University of Science and Technology in 1998. He is currently a Professor with the State Key Laboratory of Software Engineering, the School of Computer, Wuhan University and the College of Automation, Nanjing University of Posts and Telecommunications. His research interests include information security, software engineering, program analysis, machine learning and artificial intelligence

Xinge You (Senior Member, IEEE) received the B.S. and M.S. degrees in mathematics from Hubei University in 1990 and 2000, respectively, and the Ph.D. degree in computer science from the Department of Computer Science, Hong Kong Baptist University, China, in 2004. He is currently a Professor with the School of Electronic Information and Communications, Huazhong University of Science and Technology. His current research interests include computer vision, wavelet analysis and applications
Corresponding author: Sichao Fu, e-mail: fusichao_hust@hust.edu.cn; Qinmu Peng, e-mail: pengqinmu@hust.edu.cn
Received Date: 2025-03-17
Revised Date: 2025-07-06
Accepted Date: 2025-09-03

Available Online: 2026-07-22

Abstract

Abstract

Graph neural networks (GNNs) often suffer from performance degradation due to the incompleteness of node features and structure relationships in the real world. Recently emerged graph completion learning (GCL) enhances the generalization of GNNs by reconstructing the missing node features or structure relationships. Nevertheless, these proposed GCL methods are supervised by a large number of labeled nodes, which limits their applications in extremely limited labeled nodes. Moreover, the existing GCL methods either focus on feature missing or structure missing tasks, and little effort was paid to more challenging scenarios where both node features and structure relationships are simultaneously missing. In this paper, a general GCL framework with the aid of multi-level contrast graph mask autoencoders (EWS-RGCN) is proposed to improve the generalization of GNNs guided by extremely weak supervision on graphs with features and structure missing. Specifically, to alleviate the mutual interference between missing node features and structure relationships caused by message passing of GNNs, we separate the feature and structure completion into two channels. Then, a multi-level contrastive loss is introduced to simultaneously maximize the mutual information between nodes from the encoding and decoding stage, which can discover more effective supervision information from the data itself for EWS-RGCN optimization, apart from label information. To further enhance the space consistency between reconstructed node features and structure relationships, the inter-channel information cooperation module is introduced to enhance the mutual learning of feature and structure completion channels. Extensive experiments on six benchmarks demonstrate the effectiveness of our EWS-RGCN.
- Extremely weak supervision,
- features missing,
- graph neural networks (GNNs),
- node classification,
- structure missing

FullText(HTML)

Chengxiang Lei and Sichao Fu contributed equally to this work.

References(48)

References

[1]	X. Liu, M. Yan, L. Deng, G. Li, X. Ye, and D. Fan, “Sampling methods for efficient training of graph convolutional networks: A survey,” IEEE/CAA J. Autom. Sinica, vol. 9, no. 2, pp. 205–234, Feb. 2022. doi: 10.1109/JAS.2021.1004311
[2]	S. Fu, W. Liu, W. Guan, Y. Zhou, D. Tao, and C. Xu, “Dynamic graph learning convolutional networks for semi-supervised classification,” ACM Trans. Multim. Comput. Commun. Appl., vol. 17, no. 1s, Art. no. 4, Jan. 2021. doi: 10.18178/wcse.2019.06.016
[3]	Q. Zhu, Q. Xiong, Z. Yang, and Y. Yu, “RGCNU: Recurrent graph convolutional network with uncertainty estimation for remaining useful life prediction,” IEEE/CAA J. Autom. Sinica, vol. 10, no. 7, pp. 1640–1642, Jul. 2023.
[4]	S. Fu, W. Liu, K. Zhang, Y. Zhou, and D. Tao, “Semi-supervised classification by graph p-Laplacian convolutional networks,” Inform. Sci., vol. 560, pp. 92–106, Jun. 2021.
[5]	L. Bai, L. Cui, Y. Wang, M. Li, J. Li, P. S. Yu, and E. R. Hancock, “HAQJSK: Hierarchical-aligned quantum Jensen-Shannon kernels for graph classification,” IEEE Trans. Knowl. Data Eng., vol. 36, no. 11, pp. 6370–6384, Nov. 2024.
[6]	Z. Zhao, P. Wang, H. Wen, Y. Zhang, Z. Zhou, and Y. Wang, “A twist for graph classification: Optimizing causal information flow in graph neural networks,” in Proc. 38th AAAI Conf. Artificial Intelligence, Vancouver, Canada, 2024, pp. 17042−17050.
[7]	W. Liu, S. Fu, Y. Zhou, Z.-J. Zha, and L. Nie, “Human activity recognition by manifold regularization based dynamic graph convolutional networks,” Neurocomputing, vol. 444, pp. 217–225, Jul. 2021.
[8]	S. Fu, W. Liu, K. Zhang, and Y. Zhou, “Example-feature graph convolutional networks for semi-supervised classification,” Neurocomputing, vol. 461, pp. 63–76, Oct. 2021.
[9]	T. N. Kipf and M. Welling, “Semi-supervised classification with graph convolutional networks,” in Proc. 5th Int. Conf. Learning Representations, Toulon, France, 2017. [Online], Available: https://openreview.net/forum?id=SJU4ayYgl.
[10]	D. Jin, C. Huo, C. Liang, and L. Yang, “Heterogeneous graph neural network via attribute completion,” in Proc. Web Conf., Ljubljana, Slovenia, 2021, pp. 391−400.
[11]	X. Chen, S. Chen, J. Yao, H. Zheng, Y. Zhang, and I. W. Tsang, “Learning on attribute-missing graphs,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 44, no. 2, pp. 740–757, Feb. 2022.
[12]	T. He, H. Zhou, Y.-S. Ong, and G. Cong, “Not all neighbors are worth attending to: Graph selective attention networks for semi-supervised learning,” arXiv preprint arXiv: 2210.07715, 2022.
[13]	C. Lei, S. Fu, Y. Wang, W. Qiu, Y. Hu, Q. Peng, and X. You, “Self-supervised guided hypergraph feature propagation for semi-supervised classification with missing node features,” in Proc. IEEE Int. Conf. Acoustics, Speech and Signal Processing, Rhodes Island, Greece, 2023, pp. 1−5.
[14]	T. He, Y. Liu, Y.-S. Ong, X. Wu, and X. Luo, “Polarized message-passing in graph neural networks,” Artif. Intell., vol. 331, Art. no. 104129, Jun. 2024.
[15]	R. Wang, S. Mou, X. Wang, W. Xiao, Q. Ju, C. Shi, and X. Xie, “Graph structure estimation neural networks,” in Proc. Web Conf., Ljubljana, Slovenia, 2021, pp. 342−353.
[16]	L. Wu, H. Lin, Z. Liu, Z. Liu, Y. Huang, and S. Z. Li, “Homophily-enhanced self-supervision for graph structure learning: Insights and directions,” IEEE Trans. Neural Netw. Learning Syst., vol. 35, no. 9, pp. 12358–12372, Sep. 2024.
[17]	S. Fu, Q. Peng, Y. He, X. Wang, B. Zou, D. Xu, X.-Y. Jing, and X. You, “Multilevel contrastive graph masked autoencoders for unsupervised graph-structure learning,” IEEE Trans. Neural Netw. Learning Syst., vol. 36, no. 2, pp. 3464–3478, Feb. 2025.
[18]	T. He, Y.-S. Ong, and L. Bai, “Learning conjoint attentions for graph neural nets,” in Proc. 35th Int. Conf. Neural Information Processing Systems, 2021, Art. no. 202.
[19]	Y. Liu, K. Ding, J. Wang, V. Lee, H. Liu, and S. Pan, “Learning strong graph neural networks with weak information,” in Proc. 29th ACM SIGKDD Conf. Knowledge Discovery and Data Mining, Long Beach, USA, 2023, pp. 1559−1571.
[20]	C. Huo, D. Jin, Y. Li, D. He, Y.-B. Yang, and L. Wu, “T2-GNN: Graph neural networks for graphs with incomplete features and structure via teacher-student distillation,” in Proc. 37th AAAI Conf. Artificial Intelligence, Washington, USA, 2023, pp. 4339−4346.
[21]	T. H. Haveliwala, “Topic-sensitive PageRank,” in Proc. 11th Int. Conf. World Wide Web, Honolulu, USA, 2002, pp. 517−526.
[22]	M. Li, A. Micheli, Y. G. Wang, S. Pan, P. Lió, G. S. Gnecco, and M. Sanguineti, “Guest editorial: Deep neural networks for graphs: Theory, models, algorithms, and applications,” IEEE Trans. Neural Netw. Learning Syst., vol. 35, no. 4, pp. 4367–4372, Apr. 2024.
[23]	T. Wang, D. Jin, R. Wang, D. He, and Y. Huang, “Powerful graph convolutional networks with adaptive propagation mechanism for homophily and heterophily,” in Proc. 36th AAAI Conf. Artificial Intelligence, 2022, pp. 4210−4218.
[24]	J. Guo, K. Huang, X. Yi, and R. Zhang, “Graph neural networks with diverse spectral filtering,” in Proc. ACM Web Conf., Austin, USA, 2023, pp. 306−316.
[25]	M. Li, Y. Fang, Y. Wang, H. Feng, Y. Gu, L. Bai, and P. Lió, “Deep hypergraph neural networks with tight framelets,” in Proc. 39th AAAI Conf. Artificial Intelligence, Philadelphia, USA, 2025, pp. 18385−18392.
[26]	X. Gao, W. Dai, C. Li, H. Xiong, and P. Frossard, “iPool-information-based pooling in hierarchical graph neural networks,” IEEE Trans. Neural Netw. Learning Syst., vol. 33, no. 9, pp. 5032–5044, Sep. 2022.
[27]	S. Suresh, V. Budde, J. Neville, P. Li, and J. Ma, “Breaking the limit of graph neural networks by improving the assortativity of graphs with local mixing patterns,” in Proc. 27th ACM SIGKDD Conf. Knowledge Discovery & Data Mining, Singapore, 2021, pp. 1541−1551.
[28]	H. Zhou, T. He, Y.-S. Ong, G. Cong, and Q. Chen, “Differentiable clustering for graph attention,” IEEE Trans. Knowl. Data Eng., vol. 36, no. 8, pp. 3751–3764, Aug. 2024.
[29]	J. Zhang, C. Li, and Z. Zhao, “Lightweight yet efficient: An external attentive graph convolutional network with positional prompts for sequential recommendation,” ACM Trans. Inform. Syst., vol. 43, no. 3, Art. no. 79, May 2025.
[30]	Z. Luo, Y. Dong, Q. Zheng, H. Liu, and M. Luo, “Dual-channel graph contrastive learning for self-supervised graph-level representation learning,” Pattern Recognit., vol. 139, Art. no. 109448, Jul. 2023.
[31]	T. Xiao, H. Zhu, Z. Chen, and S. Wang, “Simple and asymmetric graph contrastive learning without augmentations,” in Proc. 37th Int. Conf. Neural Information Processing Systems, New Orleans, USA, 2023, Art. no. 710.
[32]	W. Dong, D. Yan, and P. Wang, “Self-supervised node representation learning via node-to-neighbourhood alignment,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 46, no. 6, pp. 4218–4233, Jun. 2024.
[33]	Z. Hou, X. Liu, Y. Cen, Y. Dong, H. Yang, C. Wang, and J. Tang, “GraphMAE: Self-supervised masked graph autoencoders,” in Proc. 28th ACM SIGKDD Conf. Knowledge Discovery and Data Mining, Washington, USA, 2022, pp. 594−604.
[34]	Q. Tan, N. Liu, X. Huang, S.-H. Choi, L. Li, R. Chen, and X. Hu, “S2GAE: Self-supervised graph autoencoders are generalizable learners with graph masking,” in Proc. 16th ACM Int. Conf. Web Search and Data Mining, Singapore, Singapore, 2023, pp. 787−795.
[35]	M. Li, Y. Zhang, S. Wang, Y. Hu, and B. Yin, “Redundancy is not what you need: An embedding fusion graph auto-encoder for self-supervised graph representation learning,” IEEE Trans. Neural Netw. Learning Syst., vol. 36, no. 2, pp. 3519–3533, Feb. 2025.
[36]	H. Taguchi, X. Liu, and T. Murata, “Graph convolutional networks for graphs containing missing features,” Future Gener. Comput. Syst., vol. 117, pp. 155–168, Apr. 2021.
[37]	A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,” in Proc. 31st Int. Conf. Neural Information Processing Systems, Long Beach, USA, 2017, pp. 6000−6010.
[38]	S. Zhang, Y. Liu, Y. Sun, and N. Shah, “Graph-less neural networks: Teaching old MLPs new tricks via distillation,” in Proc. 10th Int. Conf. Learning Representations, 2022. [Online], Available: https://iclr.cc/virtual/2022/poster/6711.
[39]	K. He, X. Chen, S. Xie, Y. Li, P. Dollár, and R. Girshick, “Masked autoencoders are scalable vision learners,” in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition, New Orleans, USA, 2022, pp. 15979−15988.
[40]	P. Sen, G. Namata, M. Bilgic, L. Getoor, B. Gallagher, and T. Eliassi-Rad, “Collective classification in network data,” AI Mag., vol. 29, no. 3, pp. 93−106, 2008.
[41]	W. Hu, M. Fey, M. Zitnik, Y. Dong, H. Ren, B. Liu, M. Catasta, and J. Leskovec, “Open graph benchmark: Datasets for machine learning on graphs,” in Proc. 34th Int. Conf. Neural Information Processing Systems, Vancouver, Canada, 2020, Art. no. 1855.
[42]	O. Shchur, M. Mumme, A. Bojchevski, and S. Günnemann, “Pitfalls of graph neural network evaluation,” arXiv preprint arXiv: 1811.05868, 2018.
[43]	W. L. Hamilton, R. Ying, and J. Leskovec, “Inductive representation learning on large graphs,” in Proc. 31st Int. Conf. Neural Information Processing Systems, Long Beach, USA, 2017, pp. 1025−1035.
[44]	P. Veličković, G. Cucurull, A. Casanova, A. Romero, P. Lió, and Y. Bengio, “Graph attention networks,” in Proc. 6th Int. Conf. Learning Representations, Vancouver, Canada, 2018. [Online], Available: https://openreview.net/forum?id=rJXMpikCZ.
[45]	J. Gasteiger, A. Bojchevski, and S. Günnemann, “Predict then propagate: Graph neural networks meet personalized PageRank,” in Proc. Int. Conf. Learning Representations, New Orleans, USA, 2019. [Online], Available: https://openreview.net/forum?id=H1gL-2A9Ym.
[46]	F. Wu, A. Souza, T. Zhang, C. Fifty, T. Yu, and K. Weinberger, “Simplifying graph convolutional networks,” in Proc. 36th Int. Conf. Machine Learning, Long Beach, USA, 2019, pp. 6861−6871.
[47]	E. Rossi, H. Kenlay, M. I. Gorinova, B. P. Chamberlain, X. Dong, and M. M. Bronstein, “On the unreasonable effectiveness of feature propagation in learning on graphs with missing node features,” in Proc. 1st Learning on Graphs Conference, 2022, pp. 1−11.
[48]	K. Ding, J. Wang, J. Caverlee, and H. Liu, “Meta propagation networks for graph few-shot semi-supervised learning,” in Proc. 36th AAAI Conf. Artificial Intelligence, 2022, pp. 6524−6531.