Extending graph neural networks to edge-level neighborhood awareness via line graph

Yutong Guo; Wenrui Guan; Qihang Guo; Yuge Wang; Keyu Liu; Xibei Yang; Yutong Guo; Wenrui Guan; Qihang Guo; Yuge Wang; Keyu Liu; Xibei Yang

doi:10.3934/era.2026040

Electronic Research Archive

2026, Volume 34, Issue 2: 866-889. doi: 10.3934/era.2026040

Previous Article Next Article

Research article

Extending graph neural networks to edge-level neighborhood awareness via line graph

1.
School of Computer, Jiangsu University of Science and Technology, Zhenjiang 212100, China
2.
School of Economics and Management, Jiangsu University of Science and Technology, Zhenjiang 212100, China

Received: 17 November 2025 Revised: 15 January 2026 Accepted: 20 January 2026 Published: 28 January 2026

Graph neural networks (GNNs) have been widely studied for handling graph-structured data. Neighborhood awareness, a mechanism for integrating context into graphs, plays a crucial role in learning node embeddings in GNNs. Existing methods typically perform neighborhood-aware steps only at the node or hop level, which limits their ability to learn edge-level semantic information. There are two main challenges in extending neighborhood awareness to the edge level: (1) designing a learning framework with message propagation of edge embeddings, which supports the construction of edge embeddings to capture pairwise node relationships and propagate messages to perceive local context, and (2) developing a fusion approach that bridges node and edge embedding spaces, thereby compressing edge-level structural information into node space to enhance the original neighborhood awareness. In this study, we propose an edge-level neighborhood awareness network (ELNA-Net) that integrates both node- and edge-level context for more comprehensive neighborhood awareness. Specifically, we use the Taylor interaction effect to construct explicit interactions between pairwise nodes, which approximates edge embeddings and captures intricate non-linear correlations. Furthermore, the adaptive edge-aware propagation module adopts a line graph to switch the roles of nodes and edges, revealing potential dependencies among neighboring edges and promoting neighborhood awareness. Finally, we propose a cross-mapping fusion approach to integrate node- and edge-level contexts within the graph convolution operator. Experimental results demonstrate that ELNA-Net achieves superior performance in semi-supervised node classification and link prediction, outperforming state-of-the-art models.
- semi-supervised learning,
- graph neural network,
- knowledge transfer,
- feature interaction
Citation: Yutong Guo, Wenrui Guan, Qihang Guo, Yuge Wang, Keyu Liu, Xibei Yang. Extending graph neural networks to edge-level neighborhood awareness via line graph[J]. Electronic Research Archive, 2026, 34(2): 866-889. doi: 10.3934/era.2026040

Related Papers:

Abstract

Graph neural networks (GNNs) have been widely studied for handling graph-structured data. Neighborhood awareness, a mechanism for integrating context into graphs, plays a crucial role in learning node embeddings in GNNs. Existing methods typically perform neighborhood-aware steps only at the node or hop level, which limits their ability to learn edge-level semantic information. There are two main challenges in extending neighborhood awareness to the edge level: (1) designing a learning framework with message propagation of edge embeddings, which supports the construction of edge embeddings to capture pairwise node relationships and propagate messages to perceive local context, and (2) developing a fusion approach that bridges node and edge embedding spaces, thereby compressing edge-level structural information into node space to enhance the original neighborhood awareness. In this study, we propose an edge-level neighborhood awareness network (ELNA-Net) that integrates both node- and edge-level context for more comprehensive neighborhood awareness. Specifically, we use the Taylor interaction effect to construct explicit interactions between pairwise nodes, which approximates edge embeddings and captures intricate non-linear correlations. Furthermore, the adaptive edge-aware propagation module adopts a line graph to switch the roles of nodes and edges, revealing potential dependencies among neighboring edges and promoting neighborhood awareness. Finally, we propose a cross-mapping fusion approach to integrate node- and edge-level contexts within the graph convolution operator. Experimental results demonstrate that ELNA-Net achieves superior performance in semi-supervised node classification and link prediction, outperforming state-of-the-art models.

References

[1]	S. Khoshraftar, A. An, A survey on graph representation learning methods, ACM Trans. Intell. Syst. Technol., 15 (2024), 1–55. https://doi.org/10.1145/3653986 doi: 10.1145/3653986
[2]	G. Xue, M. Zhong, T. Qian, J. Li, PSA-GNN: An augmented GNN framework with priori subgraph knowledge, Neural Networks, 173 (2024), 106155. https://doi.org/10.1016/j.neunet.2024.106155 doi: 10.1016/j.neunet.2024.106155
[3]	W. L. Hamilton, R. Ying, J. Leskovec, Representation learning on graphs: Methods and applications, preprint, arXiv: 1709.05584.
[4]	M. Guang, C. Yan, Y. Xu, J. Wang, C. Jiang, Graph convolutional networks with adaptive neighborhood awareness, IEEE Trans. Pattern Anal. Mach. Intell., (2024), 7392–7404. https://doi.org/10.1109/TPAMI.2024.3391356
[5]	F. Wu, A. Souza, T. Zhang, C. Fifty, T. Yu, K. Weinberger, Simplifying graph convolutional networks, preprint, arXiv: 1902.07153.
[6]	L. Peng, R. Hu, F. Kong, J. Gan, Y. Mo, X. Shi, et al., Reverse graph learning for graph neural network, IEEE Trans. Neural Networks Learn. Syst., 35 (2024), 4530–4541. https://doi.org/10.1109/TNNLS.2022.3161030 doi: 10.1109/TNNLS.2022.3161030
[7]	X. Wang, X. Yang, P. Wang, H. Yu, T. Xu, SSGCN: A sampling sequential guided graph convolutional network, Int. J. Mach. Learn. Cybern., 15 (2024), 2023–2038. https://doi.org/10.1007/s13042-023-02013-2 doi: 10.1007/s13042-023-02013-2
[8]	H. Liu, B. Yang, D. Li, Graph collaborative filtering based on dual-message propagation mechanism, IEEE Trans. Cybern., 53 (2021), 352–364. https://doi.org/10.1109/TCYB.2021.3100521 doi: 10.1109/TCYB.2021.3100521
[9]	T. N. Kipf, M. Welling, Semi-supervised classification with graph convolutional networks, preprint, arXiv: 1609.02907.
[10]	P. Veličković, G. Cucurull, A. Casanova, A. Romero, P. Liò, Y. Bengio, Graph attention networks, preprint, arXiv: 1710.10903.
[11]	H. Zeng, H. Zhou, A. Srivastava, R. Kannan, V. Prasanna, GraphSAINT: Graph sampling based inductive learning method, preprint, arXiv: 1907.04931.
[12]	Y. Ye, S. Ji, Sparse graph attention networks, IEEE Trans. Knowl. Data Eng., 35 (2021), 905–916. https://doi.org/10.1109/TKDE.2021.3072345 doi: 10.1109/TKDE.2021.3072345
[13]	Q. Guo, X. Yang, F. Zhang, T. Xu, Perturbation-augmented graph convolutional networks: A graph contrastive learning architecture for effective node classification tasks, Eng. Appl. Artif. Intell., 129 (2024), 107616. https://doi.org/10.1016/j.engappai.2023.107616 doi: 10.1016/j.engappai.2023.107616
[14]	C. Huang, M. Li, F. Cao, H. Fujita, Z. Li, X. Wu, Are graph convolutional networks with random weights feasible?, IEEE Trans. Pattern Anal. Mach. Intell., 45 (2022), 2751–2768. https://doi.org/10.1109/TPAMI.2022.3183143 doi: 10.1109/TPAMI.2022.3183143
[15]	W. Guan, X. Yang, M. Li, Q. Guo, K. Liu, Q. Sun, VQIT-GNN: A collaborative knowledge transfer for node-level structure imbalance, Pattern Recognit., 172 (2026), 112632. https://doi.org/10.1016/j.patcog.2025.112632 doi: 10.1016/j.patcog.2025.112632
[16]	Y. Yang, Y. Sun, F. Ju, S. Wang, J. Gao, B. Yin, Multi-graph fusion graph convolutional networks with pseudo-label supervision, Neural Networks, 159 (2023), 305–317. https://doi.org/10.1016/j.neunet.2022.11.027 doi: 10.1016/j.neunet.2022.11.027
[17]	Z. Chen, J. Bruna, L. Li, Supervised community detection with line graph neural networks, preprint, arXiv: 1705.08415.
[18]	J. Wang, J. Liang, J. Cui, J. Liang, Semi-supervised learning with mixed-order graph convolutional networks, Inf. Sci., 573 (2021), 171–181. https://doi.org/10.1016/j.ins.2021.05.057 doi: 10.1016/j.ins.2021.05.057
[19]	F. Hu, Y. Zhu, S. Wu, W. Huang, L. Wang, T. Tan, Graphair: Graph representation learning with neighborhood aggregation and interaction, Pattern Recognit., 112 (2021), 107745. https://doi.org/10.1016/j.patcog.2020.107745 doi: 10.1016/j.patcog.2020.107745
[20]	B. Jiang, Y. Chen, B. Wang, H. Xu, B. Luo, DropAGG: Robust graph neural networks via drop aggregation, Neural Networks, 163 (2023), 65–74. https://doi.org/10.1016/j.neunet.2023.03.022 doi: 10.1016/j.neunet.2023.03.022
[21]	L. Cai, J. Li, J. Wang, S. Ji, Line graph neural networks for link prediction, IEEE Trans. Pattern Anal. Mach. Intell., 44 (2021), 5103–5113. https://doi.org/10.1109/TPAMI.2021.3080635 doi: 10.1109/TPAMI.2021.3080635
[22]	K. Yao, J. Liang, J. Liang, M. Li, F. Cao, Multi-view graph convolutional networks with attention mechanism, Artif. Intell., 307 (2022), 103708. https://doi.org/10.1016/j.artint.2022.103708 doi: 10.1016/j.artint.2022.103708
[23]	Q. Guo, X. Yang, W. Ding, Y. Qian, Cross-graph interaction networks, IEEE Trans. Knowl. Data Eng., 173 (2025), 11059. https://doi.org/10.1109/TKDE.2025.3543377 doi: 10.1109/TKDE.2025.3543377
[24]	Q. Guo, X. Yang, M. Li, Y. Qian, Collaborative graph neural networks for augmented graphs: A local-to-global perspective, Pattern Recognit., 158 (2025), 111020. https://doi.org/10.1016/j.patcog.2024.111020 doi: 10.1016/j.patcog.2024.111020
[25]	X. Wang, S. Trajanovski, R. E. Kooij, P. Van Mieghem, Degree distribution and assortativity in line graphs of complex networks, Physica A, 445 (2016), 343–356. https://doi.org/10.1016/j.physa.2015.10.109 doi: 10.1016/j.physa.2015.10.109
[26]	Z. Zhang, S. Sun, G. Ma, C. Zhong, Line graph contrastive learning for link prediction, Pattern Recognit., 140 (2023), 109537. https://doi.org/10.1016/j.patcog.2023.109537 doi: 10.1016/j.patcog.2023.109537
[27]	X. Jiang, R. Zhu, S. Li, P. Ji, Co-embedding of nodes and edges with graph neural networks, IEEE Trans. Pattern Anal. Mach. Intell., 45 (2020), 7075–7086. https://doi.org/10.1109/TPAMI.2020.3029762 doi: 10.1109/TPAMI.2020.3029762
[28]	L. Gong, Q. Cheng, Exploiting edge features for graph neural networks, in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, (2019), 9211–9219. https://doi.org/10.1109/CVPR.2019.00943
[29]	W. Song, C. Shi, Z. Xiao, Z. Duan, Y. Xu, M. Zhang, et al., AutoInt: Automatic feature interaction learning via self-attentive neural networks, in Proceedings of the 28th ACM International Conference on Information and Knowledge Management (CIKM), (2019), 1161–1170. https://doi.org/10.1145/3357384.3357925
[30]	R. Wang, B. Fu, G. Fu, M. Wang, Deep & cross network for ad click predictions, in Proceedings of the ADKDD'17, (2017), 1–7. https://doi.org/10.1145/3124749.3124754
[31]	S. Zhu, C. Zhou, S. Pan, X. Zhu, B. Wang, Relation structure-aware heterogeneous graph neural network, in Proceedings of the 23rd International Conference on Data Mining (ICDM), (2019), 1534–1539. https://doi.org/10.1109/ICDM.2019.00203
[32]	Z. Xie, W. Zhang, B. Sheng, P. Li, C. L. Philip Chen, BaGFN: Broad attentive graph fusion network for high-order feature interactions, IEEE Trans. Neural Networks Learn. Syst., 34 (2021), 4499–4513. https://doi.org/10.1109/TNNLS.2021.3116209 doi: 10.1109/TNNLS.2021.3116209
[33]	S. Rendle, Factorization machines, in Proceedings of the 10th International Conference on Data Mining (ICDM), (2010), 995–1000. https://doi.org/10.1109/ICDM.2010.127
[34]	W. Cheng, Y. Shen, L. Huang, Adaptive factorization network: Learning adaptive-order feature interactions, in Proceedings of the AAAI Conference on Artificial Intelligence, 34 (2020), 3609–3616. https://doi.org/10.1609/aaai.v34i04.5768
[35]	Z. Zhao, Z. Yang, C. Li, Q. Zeng, W. Guan, M. Zhou, Dual feature interaction-based graph convolutional network, IEEE Trans. Knowl. Data Eng., 35 (2022), 9019–9030. https://doi.org/10.1109/TKDE.2022.3220789 doi: 10.1109/TKDE.2022.3220789
[36]	H. Deng, N. Zou, M. Du, W. Chen, G. Feng, Z. Yang, et al., Unifying fourteen post-hoc attribution methods with Taylor interactions, IEEE Trans. Pattern Anal. Mach. Intell., 46 (2024), 4625–4640. https://doi.org/10.1109/TPAMI.2024.3358410 doi: 10.1109/TPAMI.2024.3358410
[37]	T. N. Kipf, M. Welling, Variational graph auto-encoders, preprint, arXiv: 1611.07308.
[38]	A. Bojchevski, S. Günnemann, Deep Gaussian embedding of graphs: Unsupervised inductive learning via ranking, preprint, arXiv: 1707.03815.
[39]	X. Wang, H. Ji, C. Shi, B. Wang, Y. Ye, P. Cui, et al., Heterogeneous graph attention network, in Proceedings of The World Wide Web Conference (WWW), 2019. https://doi.org/10.1145/3308558.3313562
[40]	H. Pei, B. Wei, K. Chang, Y. Lei, B. Yang, Geom-GCN: Geometric graph convolutional networks, preprint, arXiv: 2002.05287.
[41]	B. Perozzi, R. Al-Rfou, S. Skiena, DeepWalk: Online learning of social representations, in Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, (2014), 701–710. https://doi.org/10.1145/2623330.2623732
[42]	J. Schulman, P. Moritz, S. Levine, M. I. Jordan, P. Abbeel, High-dimensional continuous control using generalized advantage estimation, preprint, arXiv: 1506.02438.
[43]	S. J. Ahn, M. Kim, Variational graph normalized autoencoders, in Proceedings of the 30th ACM International Conference on Information & Knowledge Management (CIKM), (2021), 2827–2831. https://doi.org/10.1145/3459637.3482215
[44]	C. R. Jack Jr, M. A. Bernstein, N. C. Fox, P. Thompson, G. Alexander, D. Harvey, et al., The Alzheimer's disease neuroimaging initiative (ADNI): MRI methods, J. Magn. Reson. Imaging, 27 (2008), 685–691. https://doi.org/10.1002/jmri.21049 doi: 10.1002/jmri.21049
[45]	V. Sanjay, P. Swarnalatha, An enhanced approach for detecting Alzheimer's disease, in Proceedings of the 2022 Smart Technologies, Communication and Robotics (STCR), (2022), 1–5. https://doi.org/10.1109/STCR55312.2022.10009274
[46]	A. Kazi, L. Cosmo, S. A. Ahmadi, N. Navab, M. M. Bronstein, Differentiable graph module (DGM) for graph convolutional networks, IEEE Trans. Pattern Anal. Mach. Intell., 45 (2022), 1606–1617. https://doi.org/10.1109/TPAMI.2022.3170249 doi: 10.1109/TPAMI.2022.3170249
[47]	F. Li, Z. Wang, Y. Guo, C. Liu, Y. Zhu, Y. Zhou, et al., Dynamic dual-graph fusion convolutional network for Alzheimer's disease diagnosis, in Proceedings of the 2023 International Conference on Image Processing (ICIP), (2023), 675–679. https://doi.org/10.1109/ICIP49359.2023.10222732

Reader Comments

Your name:*

Email:*
© 2026 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)