pith. sign in

arxiv: 2606.20283 · v1 · pith:XW6TFOSYnew · submitted 2026-06-18 · 💻 cs.LG · cs.AI

Boundary Embedding Shaping with Adaptive Contrastive Learning for Graph Structural Disentanglement

Jiaqing Chen , Zidu Yin , Yichao Cai , Yuhang Liu , Zhen Zhang , Dong Gong , Javen Qinfeng Shi This is my paper

Pith reviewed 2026-06-26 17:57 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords graph neural networkscontrastive learningstructural disentanglementboundary embeddingsnode classificationlink predictionGNN robustness
0
0 comments X

The pith

Boundary Embedding Shaping disentangles graph structures by targeting class boundaries with adaptive contrastive learning.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Graph neural networks suffer performance loss from structural entanglement, where irrelevant neighbors contaminate embeddings, and this problem is worst for nodes near class boundaries. The paper treats boundary entanglement as the main bottleneck that uniform robust methods overlook. It introduces Boundary Embedding Shaping as a plug-in module that applies adaptive contrastive learning to suppress spurious noise only in those boundary regions. The approach requires little change to model parameters yet yields higher boundary discrimination and better results on classification and link prediction.

Core claim

The paper claims that graph structural entanglement is most acute near class boundaries in the embedding space and that Boundary Embedding Shaping (BES), an adaptive contrastive learning plug-in, selectively suppresses spurious structural noise at decision boundaries, thereby improving boundary discrimination and raising accuracy in node classification and link prediction with minimal parameter perturbation.

What carries the argument

Boundary Embedding Shaping (BES): an adaptive contrastive learning plug-in module that identifies boundary regions and shapes embeddings to reduce structural noise there.

If this is right

  • BES consistently improves boundary discrimination in GNN embeddings.
  • BES outperforms existing leading robust GNN methods on standard benchmarks.
  • BES boosts GCN node classification accuracy by an average of 3.3 percent, reaching 5.0 percent on WikiCS.
  • BES delivers higher accuracy on link prediction tasks.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The boundary-focused approach could be adapted to reduce feature entanglement in non-graph settings such as image or sequence models.
  • Because BES adds little parameter change, it could be stacked with other GNN improvements without requiring full retraining.
  • Graph datasets might be pre-processed by first detecting and labeling boundary nodes to enable more targeted disentanglement methods.

Load-bearing premise

Boundary-region entanglement is the primary performance bottleneck and can be fixed by selectively suppressing noise there without harming embeddings away from the boundaries.

What would settle it

An experiment on a graph dataset where noise is distributed uniformly rather than concentrated at boundaries, measuring whether BES gains disappear when the boundary-specific targeting is disabled.

Figures

Figures reproduced from arXiv: 2606.20283 by Dong Gong, Javen Qinfeng Shi, Jiaqing Chen, Yichao Cai, Yuhang Liu, Zhen Zhang, Zidu Yin.

Figure 1
Figure 1. Figure 1: Boundary embedding shaping on WikiCS. BES structurally disentangles boundary nodes between clusters, producing sharply separable embeddings that significantly improve classification performance. hard example mining (Shrivastava et al., 2016; Xia et al., 2022) to identify boundary nodes, namely nodes predomi￾nantly surrounded by neighbors from different classes. Sub￾sequently, we selectively improve only th… view at source ↗
Figure 2
Figure 2. Figure 2: Boundary embedding shaping architecture aims to filter the irrelevant noise introduced by the variability of the local neighboring topology to obtain improved node embeddings for better classification. k-nearest neighbors), and yi is the category label of node vi . Instead of evaluating shift scores globally across the graph, we first coarsely mark the boundary region B between clusters m and n using a mar… view at source ↗
Figure 3
Figure 3. Figure 3: Analysis of boundary embedding shaping. calized noise and exhibit high computational costs on large graphs. Contrastive Representation Learning. Class-guided contrastive learning has become a cornerstone of graph rep￾resentation learning, with seminal works such as Supervised Contrastive Learning (SCL) (Khosla et al., 2020), Proto￾typical Contrastive Learning (PCL) (Li et al., 2021), and GraphECL (Xiao et … view at source ↗
Figure 4
Figure 4. Figure 4: T-SNE visualization of node embeddings generated by different graph encoders and their combinations [PITH_FULL_IMAGE:figures/full_fig_p023_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Boundary embedding shaping. (a) is for the binary scenario; (b) is for the multi-class scenario. (a) Raw Embedding (b) Boundary Embedding Shaping (c) Final Node Embeddings [PITH_FULL_IMAGE:figures/full_fig_p026_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Boundary embedding shaping for Cora. (a) Raw Embedding (b) Boundary Embedding Shaping (c) Final Node Embeddings [PITH_FULL_IMAGE:figures/full_fig_p026_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Boundary embedding shaping for CiteSeer [PITH_FULL_IMAGE:figures/full_fig_p026_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Metric convergence of node classification on different datasets during training. (a) Cora (b) CiteSeer (c) Pubmed (d) WikiCS [PITH_FULL_IMAGE:figures/full_fig_p027_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Metric convergence of link prediction on different datasets during training. 27 [PITH_FULL_IMAGE:figures/full_fig_p027_9.png] view at source ↗
read the original abstract

Graph neural networks (GNNs) excel at aggregating neighbor information for classification, yet their performance is hindered by graph structural entanglement, where spurious correlations from semantically irrelevant neighbors contaminate node embeddings. This challenge is most acute for nodes near class boundaries in the embedding space, where amplified structural noise blurs decision boundaries and destabilizes predictions. Existing robust GNN methods largely treat all nodes uniformly, ignoring boundary vulnerabilities. In this paper, to improve classification performance, we tackle graph structural disentanglement by identifying boundary-region entanglement as the primary bottleneck and propose Boundary Embedding Shaping (BES), an adaptive contrastive learning GNN plug-in module that selectively suppresses spurious structural noise at decision boundaries with minimal model parameter perturbation. Extensive experiments demonstrate that BES consistently improves boundary discrimination and outperforms existing leading methods. Notably, BES boosts GCN performance by an average of 3.3% in node classification (up to 5.0% on WikiCS) and achieves superior accuracy in link prediction.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The paper claims that graph structural entanglement is most problematic at class boundaries in GNN embeddings and proposes Boundary Embedding Shaping (BES), an adaptive contrastive learning plug-in module that selectively suppresses spurious structural noise at those boundaries with minimal parameter perturbation. It reports that BES improves boundary discrimination, boosts GCN node classification by 3.3% on average (up to 5.0% on WikiCS), and yields superior link prediction accuracy over leading methods.

Significance. If the targeted boundary treatment proves robust and generalizes beyond the reported datasets, BES could offer a lightweight, plug-in improvement for GNN robustness without uniform regularization across all nodes. The emphasis on decision-boundary vulnerabilities addresses a plausible but under-examined source of structural noise.

major comments (2)
  1. [Abstract] Abstract: the central empirical claims (3.3% average gain, 5.0% on WikiCS, superior link prediction) are stated without any reference to datasets, baselines, number of runs, or error bars, making it impossible to assess whether the reported improvements are statistically meaningful or reproducible.
  2. [Abstract] Abstract: no equations, algorithm pseudocode, or formal definition of the adaptive contrastive loss or boundary identification procedure are supplied, so the claim that BES 'selectively suppresses spurious structural noise at decision boundaries with minimal model parameter perturbation' cannot be evaluated for correctness or side effects on non-boundary nodes.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for these comments on the abstract. We respond to each point below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central empirical claims (3.3% average gain, 5.0% on WikiCS, superior link prediction) are stated without any reference to datasets, baselines, number of runs, or error bars, making it impossible to assess whether the reported improvements are statistically meaningful or reproducible.

    Authors: We agree that the abstract would be clearer with additional context. In the revised manuscript we will add a concise clause noting that results are reported on standard citation and co-authorship benchmarks (averaged over 10 runs) while keeping the abstract within length limits; full tables with baselines, run counts, and standard deviations remain in Section 4. revision: yes

  2. Referee: [Abstract] Abstract: no equations, algorithm pseudocode, or formal definition of the adaptive contrastive loss or boundary identification procedure are supplied, so the claim that BES 'selectively suppresses spurious structural noise at decision boundaries with minimal model parameter perturbation' cannot be evaluated for correctness or side effects on non-boundary nodes.

    Authors: Abstracts conventionally omit equations and pseudocode due to space constraints; the formal definitions of the boundary identification procedure and the adaptive contrastive loss appear in Section 3, together with the analysis of parameter perturbation and effects on non-boundary nodes. The abstract only summarizes the high-level contribution, which is substantiated in the main text. revision: no

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The abstract and available context contain no equations, derivations, fitted parameters presented as predictions, or self-citations that serve as load-bearing premises. The paper proposes BES as an adaptive contrastive learning module and reports empirical performance gains (e.g., 3.3% average boost on GCN) as experimental outcomes rather than results derived by construction from inputs. No self-definitional steps, renamed known results, or uniqueness theorems imported from prior author work appear. The derivation chain is therefore self-contained against external benchmarks, with the central claim resting on empirical validation.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

Review is abstract-only; no free parameters, axioms, or invented entities beyond the named module can be extracted or verified.

axioms (1)
  • domain assumption Boundary-region entanglement is the primary bottleneck for graph structural disentanglement
    Abstract states this is the challenge most acute for nodes near class boundaries.
invented entities (1)
  • Boundary Embedding Shaping (BES) module no independent evidence
    purpose: Selectively suppress spurious structural noise at decision boundaries with minimal parameter perturbation
    New module introduced in the paper; no independent evidence provided in abstract.

pith-pipeline@v0.9.1-grok · 5717 in / 1326 out tokens · 35540 ms · 2026-06-26T17:57:11.633730+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

60 extracted references

  1. [1]

    Z., van den Hengel, A., and Qi, Y

    Bao, S., Xue, Z., Chen, Q., Ou, S., Beheshti, A., Sheng, Q. Z., van den Hengel, A., and Qi, Y. Causal MVC : Causal content-style representation learning for deep multi-view clustering. In Proceedings of the International Conference on Multimedia (MM), pp.\ 1598--1606, 2025

    2025

  2. [2]

    Bishop, C. M. Latent variable models. In Learning in Graphical Models, volume 89, pp.\ 371--403. Springer, 1998

    1998

  3. [3]

    How attentive are graph attention networks? In Proceedings of the International Conference on Learning Representations (ICLR), 2022

    Brody, S., Alon, U., and Yahav, E. How attentive are graph attention networks? In Proceedings of the International Conference on Learning Representations (ICLR), 2022

    2022

  4. [4]

    Cai, Y., Liu, Y., Gao, E., Jiang, T., Zhang, Z., van den Hengel, A., and Shi, J. Q. On the value of cross-modal misalignment in multimodal representation learning. In Proceedings of the Conference on Neural Information Processing Systems (NeurIPS), 2025

    2025

  5. [5]

    Nodeimport: Imbalanced node classification with node importance assessment

    Chen, N., Liu, Z., Hooi, B., He, B., Hu, J., and Chen, J. Nodeimport: Imbalanced node classification with node importance assessment. In Proceedings of the ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), pp.\ 94--105, 2025 a

    2025

  6. [6]

    Chen, T., Kornblith, S., Norouzi, M., and Hinton, G. E. A simple framework for contrastive learning of visual representations. In Proceedings of the International Conference on Machine Learning (ICML), pp.\ 1597--1607, 2020

    2020

  7. [7]

    Improving graph contrastive learning with community structure

    Chen, X., Yue, K., Duan, L., and Yu, L. Improving graph contrastive learning with community structure. In Proceedings of the Conference on Uncertainty in Artificial Intelligence (UAI), pp.\ 568--585, 2025 b

    2025

  8. [8]

    Daunhawer, I., Bizeul, A., Palumbo, E., Marx, A., and Vogt, J. E. Identifiability results for multimodal contrastive learning. In Proceedings of the International Conference on Learning Representations (ICLR), 2023

    2023

  9. [9]

    Convolutional neural networks on graphs with fast localized spectral filtering

    Defferrard, M., Bresson, X., and Vandergheynst, P. Convolutional neural networks on graphs with fast localized spectral filtering. In Proceedings of the Conference on Neural Information Processing Systems (NeurIPS), pp.\ 3837--3845, 2016

    2016

  10. [10]

    Imagenet: A large-scale hierarchical image database

    Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., and Fei-Fei, L. Imagenet: A large-scale hierarchical image database. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp.\ 248--255, 2009

    2009

  11. [11]

    The mnist database of handwritten digit images for machine learning research

    Deng, L. The mnist database of handwritten digit images for machine learning research. IEEE Signal Processing Magazine, 29 0 (6): 0 141--142, 2012

    2012

  12. [12]

    L., Bollacker, K

    Giles, C. L., Bollacker, K. D., and Lawrence, S. Citeseer: An automatic citation indexing system. In Proceedings of the ACM International Conference on Digital Libraries (DL), pp.\ 89--98, 1998

    1998

  13. [13]

    L., Ying, Z., and Leskovec, J

    Hamilton, W. L., Ying, Z., and Leskovec, J. Inductive representation learning on large graphs. In Proceedings of the Conference on Neural Information Processing Systems (NeurIPS), pp.\ 1024--1034, 2017

    2017

  14. [14]

    Hoang, V. T. and Lee, O. Transitivity-preserving graph representation learning for bridging local connectivity and role-based similarity. In Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), pp.\ 12456--12465, 2024

    2024

  15. [15]

    Open graph benchmark: Datasets for machine learning on graphs

    Hu, W., Fey, M., Zitnik, M., Dong, Y., Ren, H., Liu, B., Catasta, M., and Leskovec, J. Open graph benchmark: Datasets for machine learning on graphs. In Proceedings of the Conference on Neural Information Processing Systems (NeurIPS), 2020

    2020

  16. [16]

    Supervised contrastive learning

    Khosla, P., Teterwak, P., Wang, C., Sarna, A., Tian, Y., Isola, P., Maschinot, A., Liu, C., and Krishnan, D. Supervised contrastive learning. In Proceedings of the Conference on Neural Information Processing Systems (NeurIPS), 2020

    2020

  17. [17]

    Kipf, T. N. and Welling, M. Semi-supervised classification with graph convolutional networks. In Proceedings of the International Conference on Learning Representations (ICLR), 2017

    2017

  18. [18]

    Koh, P. W. and Liang, P. Understanding black-box predictions via influence functions. In Proceedings of the International Conference on Machine Learning (ICML), pp.\ 1885--1894, 2017

    2017

  19. [19]

    Li, J., Zhou, P., Xiong, C., and Hoi, S. C. Prototypical contrastive learning of unsupervised representations. In Proceedings of the International Conference on Learning Representations (ICLR), 2021

    2021

  20. [20]

    Kolmogorov--arnold graph neural networks for molecular property prediction

    Li, L., Zhang, Y., Wang, G., and Xia, K. Kolmogorov--arnold graph neural networks for molecular property prediction. Nature Machine Intelligence, 7 0 (8): 0 1346--1354, 2025

    2025

  21. [21]

    Liu, Y., Zhang, Z., Gong, D., Gong, M., Huang, B., van den Hengel, A., Zhang, K., and Shi, J. Q. Identifiable latent polynomial causal models through the lens of change. In Proceedings of the International Conference on Learning Representations (ICLR), 2024

    2024

  22. [22]

    Liu, Y., Zhang, Z., Gong, D., Gong, M., Huang, B., van den Hengel, A., Zhang, K., and Shi, J. Q. Latent covariate shift: Unlocking partial identifiability for multi-source domain adaptation. Transactions on Machine Learning Research, 2025. ISSN 2835-8856

    2025

  23. [23]

    Liu, Y., Gong, D., Cai, Y., Gao, E., Zhang, Z., Huang, B., Gong, M., van den Hengel, A., and Shi, J. Q. I predict therefore i am: Is next token prediction enough to learn human-interpretable concepts from data? In Proceedings of the International Conference on Learning Representations (ICLR), 2026 a

    2026

  24. [24]

    Liu, Y., Zhang, Z., Gong, D., Gao, E., Huang, B., Gong, M., van den Hengel, A., Zhang, K., and Shi, J. Q. Beyond DAG s: A latent partial causal model for multimodal learning. In Proceedings of the International Conference on Learning Representations (ICLR), 2026 b

    2026

  25. [25]

    Liu, Y., Zhang, Z., Gong, D., Gong, M., Huang, B., van den Hengel, A., Zhang, K., and Shi, J. Q. Identifying weight-variant latent causal models. Journal of Machine Learning Research, 27 0 (4): 0 1--49, 2026 c

    2026

  26. [26]

    Classic gnns are strong baselines: Reassessing gnns for node classification

    Luo, Y., Shi, L., and Wu, X.-M. Classic gnns are strong baselines: Reassessing gnns for node classification. In Proceedings of the Conference on Neural Information Processing Systems (NeurIPS), pp.\ 97650--97669, 2024

    2024

  27. [27]

    Enhancing robustness of graph neural networks on social media with explainable inverse reinforcement learning

    Lyu, Y., Li, C., Xie, S., and Zhang, X. Enhancing robustness of graph neural networks on social media with explainable inverse reinforcement learning. In Proceedings of the Conference on Neural Information Processing Systems (NeurIPS), pp.\ 31736--31758, 2024

    2024

  28. [28]

    Disentangled graph convolutional networks

    Ma, J., Cui, P., Kuang, K., Wang, X., and Zhu, W. Disentangled graph convolutional networks. In Proceedings of the International Conference on Machine Learning (ICML), pp.\ 4212--4221, 2019

    2019

  29. [29]

    K., Nigam, K., Rennie, J., and Seymore, K

    McCallum, A. K., Nigam, K., Rennie, J., and Seymore, K. Automating the construction of internet portals with machine learning. Information Retrieval, 3: 0 127--163, 2000

    2000

  30. [30]

    and Cangea, C

    Mernyei, P. and Cangea, C. Wiki-cs: A wikipedia-based benchmark for graph neural networks. arXiv preprint arXiv:2007.02901, 2020

    arXiv 2007

  31. [31]

    Interpretable and generalizable graph learning via stochastic attention mechanism

    Miao, S., Liu, M., and Li, P. Interpretable and generalizable graph learning via stochastic attention mechanism. In Proceedings of the International Conference on Machine Learning (ICML), pp.\ 15524--15543, 2022

    2022

  32. [32]

    L., Lenssen, J

    Morris, C., Ritzert, M., Fey, M., Hamilton, W. L., Lenssen, J. E., Rattan, G., and Grohe, M. Weisfeiler and leman go neural: Higher-order graph neural networks. In Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), pp.\ 4602--4609, 2019

    2019

  33. [33]

    C.-C., Lei, Y., and Yang, B

    Pei, H., Wei, B., Chang, K. C.-C., Lei, Y., and Yang, B. Geom-gcn: Geometric graph convolutional networks. In Proceedings of the International Conference on Learning Representations (ICLR), 2020

    2020

  34. [34]

    Pennington, J., Socher, R., and Manning, C. D. Glove: Global vectors for word representation. In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP), pp.\ 1532--1543, 2014

    2014

  35. [35]

    A critical look at the evaluation of gnns under heterophily: Are we really making progress? In Proceedings of the International Conference on Learning Representations (ICLR), 2023

    Platonov, O., Kuznedelev, D., Diskin, M., Babenko, A., and Prokhorenkova, L. A critical look at the evaluation of gnns under heterophily: Are we really making progress? In Proceedings of the International Conference on Learning Representations (ICLR), 2023

    2023

  36. [36]

    K., Bronstein, M

    Rusch, T. K., Bronstein, M. M., and Mishra, S. A survey on oversmoothing in graph neural networks. arXiv preprint arXiv:2303.10993, 2023

    arXiv 2023

  37. [37]

    Collective classification in network data

    Sen, P., Namata, G., Bilgic, M., Getoor, L., Galligher, B., and Eliassi-Rad, T. Collective classification in network data. AI Magazine, 29 0 (3): 0 93--106, 2008

    2008

  38. [38]

    Training region-based object detectors with online hard example mining

    Shrivastava, A., Gupta, A., and Girshick, R. Training region-based object detectors with online hard example mining. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp.\ 761--769, 2016

    2016

  39. [39]

    Networked inequality: Preferential attachment bias in graph neural network link prediction

    Subramonian, A., Sagun, L., and Sun, Y. Networked inequality: Preferential attachment bias in graph neural network link prediction. In Proceedings of the International Conference on Machine Learning (ICML), pp.\ 46891--46925, 2024

    2024

  40. [40]

    Sun, L., Huang, Z., Wang, Z., Wang, F., Peng, H., and Yu, P. S. Motif-aware riemannian graph neural network with generative-contrastive learning. In Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), pp.\ 9044--9052, 2024

    2024

  41. [41]

    Graph attention networks

    Velickovic, P., Cucurull, G., Casanova, A., Romero, A., Li \` o , P., and Bengio, Y. Graph attention networks. In Proceedings of the International Conference on Learning Representations (ICLR), 2018

    2018

  42. [42]

    u gelgen, J., Sharma, Y., Gresele, L., Brendel, W., Sch \

    von K \" u gelgen, J., Sharma, Y., Gresele, L., Brendel, W., Sch \" o lkopf, B., Besserve, M., and Locatello, F. Self-supervised learning with data augmentations provably isolates content from style. In Proceedings of the Conference on Neural Information Processing Systems (NeurIPS), pp.\ 16451--16467, 2021

    2021

  43. [43]

    Generalizing aggregation functions in gnns: Building high capacity and robust gnns via nonlinear aggregation

    Wang, B., Jiang, B., Tang, J., and Luo, B. Generalizing aggregation functions in gnns: Building high capacity and robust gnns via nonlinear aggregation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45 0 (11): 0 13454--13466, 2023

    2023

  44. [44]

    Disentangled graph collaborative filtering

    Wang, X., Jin, H., Zhang, A., He, X., Xu, T., and Chua, T. Disentangled graph collaborative filtering. In Proceedings of the ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), pp.\ 1001--1010, 2020

    2020

  45. [45]

    Simplifying graph convolutional networks

    Wu, F., Souza, A., Zhang, T., Fifty, C., Yu, T., and Weinberger, K. Simplifying graph convolutional networks. In Proceedings of the International Conference on Machine Learning (ICML), pp.\ 6861--6871, 2019

    2019

  46. [46]

    Graph neural networks in recommender systems: A survey

    Wu, S., Sun, F., Zhang, W., Xie, X., and Cui, B. Graph neural networks in recommender systems: A survey. ACM Computing Surveys, 55 0 (5): 0 1--37, 2022

    2022

  47. [47]

    Xia, J., Wu, L., Wang, G., Chen, J., and Li, S. Z. Pro GCL : Rethinking hard negative mining in graph contrastive learning. In Proceedings of the International Conference on Machine Learning (ICML), pp.\ 24332--24346, 2022

    2022

  48. [48]

    C., Wang, S., and Honavar, V

    Xiao, T., Zhu, H., Zhang, Z., Guo, Z., Aggarwal, C. C., Wang, S., and Honavar, V. G. Efficient contrastive learning for fast and accurate inference on graphs. In Proceedings of the International Conference on Machine Learning (ICML), pp.\ 54363--54381, 2024

    2024

  49. [49]

    Revisiting semi-supervised learning with graph embeddings

    Yang, Z., Cohen, W., and Salakhudinov, R. Revisiting semi-supervised learning with graph embeddings. In Proceedings of the International Conference on Machine Learning (ICML), pp.\ 40--48, 2016

    2016

  50. [50]

    v., and Locatello, F

    Yao, D., Xu, D., Lachapelle, S., Magliacane, S., Taslakian, P., Martius, G., K \"u gelgen, J. v., and Locatello, F. Multi-view causal representation learning with partial observability. In Proceedings of the International Conference on Learning Representations (ICLR), 2024

    2024

  51. [51]

    Do transformers really perform badly for graph representation? In Proceedings of the Conference on Neural Information Processing Systems (NeurIPS), pp.\ 28877--28888, 2021

    Ying, C., Cai, T., Luo, S., Zheng, S., Ke, G., He, D., Shen, Y., and Liu, T. Do transformers really perform badly for graph representation? In Proceedings of the Conference on Neural Information Processing Systems (NeurIPS), pp.\ 28877--28888, 2021

    2021

  52. [52]

    Graph contrastive learning with augmentations

    You, Y., Chen, T., Sui, Y., Chen, T., Wang, Z., and Shen, Y. Graph contrastive learning with augmentations. In Proceedings of the Conference on Neural Information Processing Systems (NeurIPS), pp.\ 5812--5823, 2020

    2020

  53. [53]

    Zeng, H., Zhou, H., Srivastava, A., Kannan, R., and Prasanna, V. K. Graph SAINT : Graph sampling based inductive learning method. In Proceedings of the International Conference on Learning Representations (ICLR), 2020

    2020

  54. [54]

    Decouple graph neural networks: Train multiple simple gnns simultaneously instead of one

    Zhang, H., Zhu, Y., and Li, X. Decouple graph neural networks: Train multiple simple gnns simultaneously instead of one. IEEE Transactions on Pattern Analysis and Machine Intelligence, 46 0 (11): 0 7451--7462, 2024 a

    2024

  55. [55]

    Graph neural networks in modern ai-aided drug discovery

    Zhang, O., Lin, H., Zhang, X., Wang, X., Wu, Z., Ye, Q., Zhao, W., Wang, J., Ying, K., Kang, Y., et al. Graph neural networks in modern ai-aided drug discovery. Chemical Reviews, 125 0 (20): 0 10001--10103, 2025

    2025

  56. [56]

    Trans GNN : Harnessing the collaborative power of transformers and graph neural networks for recommender systems

    Zhang, P., Yan, Y., Zhang, X., Li, C., Wang, S., Huang, F., and Kim, S. Trans GNN : Harnessing the collaborative power of transformers and graph neural networks for recommender systems. In Proceedings of the ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), pp.\ 1285--1295, 2024 b

    2024

  57. [57]

    and Koniusz, P

    Zhu, H. and Koniusz, P. Simple spectral graph convolution. In Proceedings of the International Conference on Learning Representations (ICLR), 2021

    2021

  58. [58]

    Beyond homophily in graph neural networks: Current limitations and effective designs

    Zhu, J., Yan, Y., Zhao, L., Heimann, M., Akoglu, L., and Koutra, D. Beyond homophily in graph neural networks: Current limitations and effective designs. In Proceedings of the Conference on Neural Information Processing Systems (NeurIPS), pp.\ 7793--7804, 2020

    2020

  59. [59]

    Graph contrastive learning with adaptive augmentation

    Zhu, Y., Xu, Y., Yu, F., Liu, Q., Wu, S., and Wang, L. Graph contrastive learning with adaptive augmentation. In Proceedings of the Web Conference (WWW), pp.\ 2069--2080, 2021

    2069

  60. [60]

    S., Sharma, Y., Schneider, S., Bethge, M., and Brendel, W

    Zimmermann, R. S., Sharma, Y., Schneider, S., Bethge, M., and Brendel, W. Contrastive learning inverts the data generating process. In Proceedings of the International Conference on Machine Learning (ICML), pp.\ 12979--12990, 2021

    2021