pith. sign in

arxiv: 2406.15797 · v2 · pith:66W6QL56new · submitted 2024-06-22 · 💻 cs.LG · cs.AI

texttt{SynC}: Synergistic Boosting of Structure and Representation for Deep Graph Clustering

Shifei Ding , Benyu Wu , Xiao Xu , Ling Ding , Xindong Wu This is my paper

Pith reviewed 2026-05-24 00:12 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords deep graph clusteringgraph auto-encoderstructure augmentationrepresentation learningself-supervised clusteringlow homophily graphsgraph neural networks
0
0 comments X

The pith

SynC improves graph clustering by alternating embedding learning and structure refinement in weight-shared stages that reinforce each other.

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

The paper claims that representation learning and structure augmentation in graphs form a reciprocal loop where better embeddings enable more reliable graph changes and vice versa. It introduces a two-stage process in which a Transform Input Graph Auto-Encoder first produces collapse-resistant embeddings to guide structure augmentation, after which neighborhood representations are re-learned on the updated graph for self-supervised clustering. The stages share weights so that each pass strengthens the other while cutting total parameters. A separate fine-tuning step is added to maintain performance when the input graph has low homophily.

Core claim

SynC employs a Transform Input Graph Auto-Encoder (TIGAE) to obtain high-quality embeddings via mitigating the representations collapse issue of GAE for guiding structure augmentation. Then, we re-capture neighborhood representations on the refined graph to obtain clustering-friendly embeddings and conduct self-supervised clustering. Notably, these two stages share weights, resulting in synergistic boosting while significantly reducing the number of model parameters. Additionally, we introduce a structure fine-tuning strategy to improve the model's generalization on the low homophily graph.

What carries the argument

The weight-shared alternation between TIGAE-based embedding and structure augmentation on the refined graph, followed by self-supervised clustering.

If this is right

  • More cohesive node representations lead to more reliable structure augmentation on subsequent passes.
  • Shared weights across stages reduce the total number of trainable parameters compared with separate models.
  • The fine-tuning strategy extends usable performance to graphs where connected nodes are not strongly similar.
  • Self-supervised clustering on the final embeddings yields higher accuracy than methods that treat representation and structure separately.

Where Pith is reading between the lines

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

  • The same alternating shared-weight pattern could be tested on graph tasks other than clustering, such as link prediction.
  • If the reciprocal loop holds, removing the structure-augmentation step entirely should produce a measurable drop in final clustering quality.
  • The approach may require fewer total training epochs than two independent models because each stage supplies a better starting point for the next.

Load-bearing premise

The embeddings produced by the first stage are accurate enough to guide structure changes that genuinely improve the second stage rather than introducing errors that accumulate.

What would settle it

An ablation experiment in which the two stages run without weight sharing and produce equal or better clustering accuracy than the shared-weight version would falsify the synergistic-boosting claim.

Figures

Figures reproduced from arXiv: 2406.15797 by Benyu Wu, Ling Ding, Shifei Ding, Xiao Xu, Xindong Wu.

Figure 1
Figure 1. Figure 1: Comparison of nodes similarity on the dataset ACM. The first row [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of our proposed SynC framework. The TIGAE combines linear transformation with graph convolutional networks. Firstly, we employ the [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Ablation results of the proposed SynC on four datasets. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Clustering results with different hyper-parameters. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Convergence effect on different datasets. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: 2D t-SNE visualization of eight methods. From the first row to the third row, the datasets are ACM, CITE, DBLP. increase in the loss value and a decline in the accuracy. It can be primarily caused by the loss used in pre-training, which differs from training, so using pre-trained parameters has a lower loss in the initial stage. However, as we modified the graph and re-aggregated the features, the discrimi… view at source ↗
Figure 8
Figure 8. Figure 8: Pre-training and training time. Our proposed method requires pre-training first and then fine-tuning. However, experiments on the five datasets in [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
read the original abstract

Employing graph neural networks (GNNs) for graph clustering has shown promising results in deep graph clustering. However, existing methods disregard the reciprocal relationship between representation learning and structure augmentation: the more homogeneous the graph, the more cohesive the node representations; the more cohesive the node representations, the more reliable the structure augmentation becomes. Moreover, the generalization ability of existing GNN-based models on the low homophily graph is relatively poor. To this end, we propose a graph clustering framework named Synergistic Deep Graph Clustering Network (SynC). SynC employs a Transform Input Graph Auto-Encoder (TIGAE) to obtain high-quality embeddings via mitigating the representations collapse issue of GAE for guiding structure augmentation. Then, we re-capture neighborhood representations on the refined graph to obtain clustering-friendly embeddings and conduct self-supervised clustering. Notably, these two stages share weights, resulting in synergistic boosting while significantly reducing the number of model parameters. Additionally, we introduce a structure fine-tuning strategy to improve the model's generalization on the low homophily graph. Extensive experiments on benchmark datasets demonstrate the superiority of SynC. The code is released at GitHub.

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 / 1 minor

Summary. The paper proposes SynC, a deep graph clustering framework that exploits the reciprocal relationship between representation learning and structure augmentation. It introduces a Transform Input Graph Auto-Encoder (TIGAE) to obtain high-quality embeddings by mitigating the representation collapse issue of standard GAE, using these to guide structure augmentation. Neighborhood representations are then re-captured on the refined graph to produce clustering-friendly embeddings for self-supervised clustering. The two stages share weights to achieve synergistic boosting and parameter reduction; a structure fine-tuning strategy is added to improve generalization on low-homophily graphs. Experiments on benchmark datasets are claimed to demonstrate superiority.

Significance. If the TIGAE mechanism reliably mitigates collapse to enable effective structure augmentation, and if shared weights produce genuine synergistic improvement beyond alternation plus parameter savings, the framework could advance GNN-based clustering by providing a more efficient, reciprocal-learning approach with better low-homophily generalization. The released code supports reproducibility.

major comments (2)
  1. [Abstract] Abstract: the central claims of synergistic boosting via shared weights and collapse mitigation by TIGAE are asserted without any equations, pseudocode, ablation design, or quantitative evidence; this prevents verification of whether the reciprocal relationship produces more than simple alternation or added regularization.
  2. [Abstract] Abstract: the structure fine-tuning strategy for low-homophily graphs is described only at a high level with no details on its formulation, how it differs from standard augmentation, or its interaction with the shared-weight stages.
minor comments (1)
  1. [Abstract] The GitHub code release is a positive step for reproducibility, but the abstract provides no link or repository details.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed review and the opportunity to clarify aspects of our work. Below we respond point-by-point to the major comments on the abstract. The full manuscript provides the requested technical details, equations, and experimental evidence in the body of the paper.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claims of synergistic boosting via shared weights and collapse mitigation by TIGAE are asserted without any equations, pseudocode, ablation design, or quantitative evidence; this prevents verification of whether the reciprocal relationship produces more than simple alternation or added regularization.

    Authors: We agree that the abstract is a high-level summary and does not contain equations or ablations. The TIGAE formulation, including the transformation that mitigates representation collapse, appears with equations in Section 3.2. The shared-weight synergistic mechanism is formalized in Section 3.3. Ablation studies isolating the reciprocal boosting effect (beyond alternation or regularization) are reported in Section 4.4 with quantitative results in Tables 2–4. If the editor prefers, we can add a single sentence to the abstract referencing these sections while remaining within length limits. revision: partial

  2. Referee: [Abstract] Abstract: the structure fine-tuning strategy for low-homophily graphs is described only at a high level with no details on its formulation, how it differs from standard augmentation, or its interaction with the shared-weight stages.

    Authors: The abstract summarizes the fine-tuning strategy at a high level, as is conventional. Its precise formulation (an embedding-guided edge refinement step), differences from the main augmentation pipeline, and integration with the shared-weight TIGAE stages are specified in Section 3.4. Experiments on low-homophily graphs (Tables 5–6) demonstrate its contribution. We are willing to insert one clarifying clause in the abstract if space permits. revision: partial

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper proposes an architectural framework (TIGAE for embeddings guiding structure augmentation, followed by re-capture on refined graph with shared weights) motivated by a claimed reciprocal relationship, but presents this as an iterative design choice rather than a mathematical derivation. No equations, fitting procedures, or self-citations are described that reduce any claimed prediction or result to its inputs by construction. The method is self-contained as a proposed clustering network with external experimental validation on benchmarks, and the shared-weights synergy is asserted as an empirical outcome rather than a definitional tautology.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no concrete free parameters, axioms, or invented entities; the method description does not expose fitted constants, unproven lemmas, or new postulated objects.

pith-pipeline@v0.9.0 · 5748 in / 1219 out tokens · 20525 ms · 2026-05-24T00:12:01.041998+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

46 extracted references · 46 canonical work pages · 1 internal anchor

  1. [1]

    Attributed graph clustering: A deep attentional embedding approach,

    C. Wang, S. Pan, R. Hu, G. Long, J. Jiang, and C. Zhang, “Attributed graph clustering: A deep attentional embedding approach,” in Proc. of IJCAI. Macao, China: ijcai.org, 2019, pp. 3670–3676

    work page 2019

  2. [2]

    Attention-driven graph clustering network,

    Z. Peng, H. Liu, Y . Jia, and J. Hou, “Attention-driven graph clustering network,” in Proc. of ACM MM . Virtual Event, China: ACM, 2021, pp. 935–943

    work page 2021

  3. [3]

    Attributed graph clustering with dual redundancy reduction,

    L. Gong, S. Zhou, W. Tu, and X. Liu, “Attributed graph clustering with dual redundancy reduction,” in Proc. of IJCAI . Vienna, Austria: ijcai.org, 2022, pp. 3015–3021

    work page 2022

  4. [4]

    Deep graph clustering with enhanced feature representations for community detection,

    J. Hao and W. Zhu, “Deep graph clustering with enhanced feature representations for community detection,” Applied Intelligence, vol. 53, no. 2, pp. 1336–1349, 2023

    work page 2023

  5. [5]

    Hard sample aware network for contrastive deep graph clustering,

    Y . Liu, X. Yang, S. Zhou, X. Liu, Z. Wang, K. Liang, W. Tu, L. Li, J. Duan, and C. Chen, “Hard sample aware network for contrastive deep graph clustering,” in Proc. of AAAI . Washington, DC, USA: AAAI Press, 2023, pp. 8914–8922

    work page 2023

  6. [6]

    Graph clustering network with structure embedding enhanced,

    S. Ding, B. Wu, X. Xu, L. Guo, and L. Ding, “Graph clustering network with structure embedding enhanced,” Pattern Recognit. , vol. 144, p. 109833, 2023

    work page 2023

  7. [7]

    Semi-supervised classification with graph convolutional networks,

    T. N. Kipf and M. Welling, “Semi-supervised classification with graph convolutional networks,” 2016

    work page 2016

  8. [8]

    Variational Graph Auto-Encoders

    ——, “Variational graph auto-encoders,” CoRR, vol. abs/1611.07308, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  9. [9]

    Deep graph clustering via dual correlation reduction,

    Y . Liu, W. Tu, S. Zhou, X. Liu, L. Song, X. Yang, and E. Zhu, “Deep graph clustering via dual correlation reduction,” in Proc. of AAAI. Virtual Event: AAAI Press, 2022, pp. 7603–7611

    work page 2022

  10. [10]

    Deeper insights into graph convolutional networks for semi-supervised learning,

    Q. Li, Z. Han, and X. Wu, “Deeper insights into graph convolutional networks for semi-supervised learning,” inProc. of AAAI. New Orleans, Louisiana, USA: AAAI Press, 2018, pp. 3538–3545

    work page 2018

  11. [11]

    Data augmentation for graph neural networks,

    T. Zhao, Y . Liu, L. Neves, O. J. Woodford, M. Jiang, and N. Shah, “Data augmentation for graph neural networks,” in Proc. of AAAI . Virtual Event: AAAI Press, 2021, pp. 11 015–11 023

    work page 2021

  12. [12]

    The concrete distribution: A continuous relaxation of discrete random variables,

    C. J. Maddison, A. Mnih, and Y . W. Teh, “The concrete distribution: A continuous relaxation of discrete random variables,” in International Conference on Learning Representations , 2017. [Online]. Available: https://openreview.net/forum?id=S1jE5L5gl

    work page 2017

  13. [13]

    Algorithm as 136: A k-means clustering algorithm,

    J. Hartigan and M. Wong, “Algorithm as 136: A k-means clustering algorithm,” Journal of the royal statistical society. series c (applied statistics), vol. 28, no. 1, pp. 100–108, 1979

    work page 1979

  14. [14]

    Unsupervised deep embedding for clustering analysis,

    J. Xie, R. B. Girshick, and A. Farhadi, “Unsupervised deep embedding for clustering analysis,” in Proc. of ICML , ser. JMLR Workshop and Conference Proceedings, vol. 48. New York, NY: JMLR.org, 2016, pp. 478–487

    work page 2016

  15. [15]

    Improved deep embedded clustering with local structure preservation,

    X. Guo, L. Gao, X. Liu, and J. Yin, “Improved deep embedded clustering with local structure preservation,” in Proc. of IJCAI . Melbourne, Australia: ijcai.org, 2017, pp. 1753–1759

    work page 2017

  16. [16]

    Graph attention networks,

    P. Velickovic, G. Cucurull, A. Casanova, A. Romero, P. Li `o, and Y . Bengio, “Graph attention networks,” in Proc. of ICLR . Vancouver, BC, Canada: OpenReview.net, 2018

    work page 2018

  17. [17]

    Adversarially regularized graph autoencoder for graph embedding,

    S. Pan, R. Hu, G. Long, J. Jiang, L. Yao, and C. Zhang, “Adversarially regularized graph autoencoder for graph embedding,” in Proc. of IJCAI. Stockholm, Sweden: ijcai.org, 2018, pp. 2609–2615

    work page 2018

  18. [18]

    MGAE: marginalized graph autoencoder for graph clustering,

    C. Wang, S. Pan, G. Long, X. Zhu, and J. Jiang, “MGAE: marginalized graph autoencoder for graph clustering,” in Proc. of CIKM. Singapore: ACM, 2017, pp. 889–898

    work page 2017

  19. [19]

    Rethink- ing graph auto-encoder models for attributed graph clustering,

    N. Mrabah, M. Bouguessa, M. F. Touati, and R. Ksantini, “Rethink- ing graph auto-encoder models for attributed graph clustering,” IEEE Transactions on Knowledge and Data Engineering , vol. 35, no. 9, pp. 9037–9053, 2023

    work page 2023

  20. [20]

    What’s behind the mask: Understanding masked graph modeling for graph autoencoders,

    J. Li, R. Wu, W. Sun, L. Chen, S. Tian, L. Zhu, C. Meng, Z. Zheng, and W. Wang, “What’s behind the mask: Understanding masked graph modeling for graph autoencoders,” in Proc. of KDD , ser. KDD ’23. New York, NY , USA: Association for Computing Machinery, 2023, p. 1268–1279

    work page 2023

  21. [21]

    Graphmae2: A decoding-enhanced masked self-supervised graph learner,

    Z. Hou, Y . He, Y . Cen, X. Liu, Y . Dong, E. Kharlamov, and J. Tang, “Graphmae2: A decoding-enhanced masked self-supervised graph learner,” in Proc. of WWW , ser. WWW ’23. New York, NY , USA: Association for Computing Machinery, 2023, p. 737–746

    work page 2023

  22. [22]

    Structural deep clustering network,

    D. Bo, X. Wang, C. Shi, M. Zhu, E. Lu, and P. Cui, “Structural deep clustering network,” in Proc. of WWW. Taipei, Taiwan: ACM / IW3C2, 2020, pp. 1400–1410

    work page 2020

  23. [23]

    Deep fusion clustering network,

    W. Tu, S. Zhou, X. Liu, X. Guo, Z. Cai, E. Zhu, and J. Cheng, “Deep fusion clustering network,” in Proc. of AAAI . Virtual Event: AAAI Press, 2021, pp. 9978–9987

    work page 2021

  24. [24]

    Caegcn: Cross-attention fusion based enhanced graph convolutional network for clustering,

    G. Huo, Y . Zhang, J. Gao, B. Wang, Y . Hu, and B. Yin, “Caegcn: Cross-attention fusion based enhanced graph convolutional network for clustering,” IEEE Transactions on Knowledge and Data Engineering , vol. 35, no. 4, pp. 3471–3483, 2021

    work page 2021

  25. [25]

    Adaptive graph convolution methods for attributed graph clustering,

    X. Zhang, H. Liu, Q. Li, X.-M. Wu, and X. Zhang, “Adaptive graph convolution methods for attributed graph clustering,” IEEE Transactions on Knowledge and Data Engineering, vol. 35, no. 12, pp. 12 384–12 399, 2023

    work page 2023

  26. [26]

    Redundancy-free self-supervised relational learning for graph cluster- ing,

    S. Yi, W. Ju, Y . Qin, X. Luo, L. Liu, Y . Zhou, and M. Zhang, “Redundancy-free self-supervised relational learning for graph cluster- ing,” IEEE Transactions on Neural Networks and Learning Systems , 2023

    work page 2023

  27. [27]

    Graph debiased contrastive learning with joint representation clustering,

    H. Zhao, X. Yang, Z. Wang, E. Yang, and C. Deng, “Graph debiased contrastive learning with joint representation clustering,” in Proc. of IJCAI. Montreal, Canada: ijcai.org, 2021, pp. 3434–3440

    work page 2021

  28. [28]

    Cluster-guided contrastive graph clustering network,

    X. Yang, Y . Liu, S. Zhou, S. Wang, W. Tu, Q. Zheng, X. Liu, L. Fang, and E. Zhu, “Cluster-guided contrastive graph clustering network,” in Proc. of AAAI. Washington, DC, USA: AAAI Press, 2023, pp. 10 834– 10 842

    work page 2023

  29. [29]

    Simple contrastive graph clustering,

    Y . Liu, X. Yang, S. Zhou, X. Liu, S. Wang, K. Liang, W. Tu, and L. Li, “Simple contrastive graph clustering,” IEEE Transactions on Neural Networks and Learning Systems , pp. 1–12, 2023

    work page 2023

  30. [30]

    Graph learning for multiview clustering,

    K. Zhan, C. Zhang, J. Guan, and J. Wang, “Graph learning for multiview clustering,” IEEE Transactions on Cybernetics , vol. 48, no. 10, pp. 2887–2895, 2018

    work page 2018

  31. [31]

    Stationary diffusion state neural estimation for multiview clustering,

    C. Liu, Z. Liao, Y . Ma, and K. Zhan, “Stationary diffusion state neural estimation for multiview clustering,” in Proc. of AAAI . Virtual Event: AAAI Press, 2022, pp. 7542–7549

    work page 2022

  32. [32]

    View-consistent heterogeneous network on graphs with few labeled nodes,

    Z. Liao, X. Zhang, W. Su, and K. Zhan, “View-consistent heterogeneous network on graphs with few labeled nodes,” IEEE Trans. Cybern. , vol. 53, no. 9, pp. 5523–5532, 2023

    work page 2023

  33. [33]

    Dink-net: Neural clustering on large graphs,

    Y . Liu, K. Liang, J. Xia, S. Zhou, X. Yang, X. Liu, and S. Z. Li, “Dink-net: Neural clustering on large graphs,” in Proc. of ICML , ser. Proceedings of Machine Learning Research, vol. 202. Honolulu, Hawaii, USA: PMLR, 2023, pp. 21 794–21 812

    work page 2023

  34. [34]

    Reinforcement graph clustering with unknown cluster number,

    Y . Liu, K. Liang, J. Xia, X. Yang, S. Zhou, M. Liu, X. Liu, and S. Z. Li, “Reinforcement graph clustering with unknown cluster number,” in Proc. of ACM MM , ser. MM ’23. New York, NY , USA: Association for Computing Machinery, 2023, p. 3528–3537

    work page 2023

  35. [35]

    A survey of deep graph clustering: Taxonomy, challenge, application, and open resource,

    Y . Liu, J. Xia, S. Zhou, X. Yang, K. Liang, C. Fan, Y . Zhuang, S. Z. Li, X. Liu, and K. He, “A survey of deep graph clustering: Taxonomy, challenge, application, and open resource,” 2023

    work page 2023

  36. [36]

    Graph augmentation learning,

    S. Yu, H. Huang, M. N. Dao, and F. Xia, “Graph augmentation learning,” in Proc. of WWW . Lyon, France: ACM, 2022, pp. 1063–1072

    work page 2022

  37. [37]

    Dropgnn: Random dropouts increase the expressiveness of graph neural networks,

    P. A. Papp, K. Martinkus, L. Faber, and R. Wattenhofer, “Dropgnn: Random dropouts increase the expressiveness of graph neural networks,” in Proc. of NeurIPS . virtual: MIT Press, 2021, pp. 21 997–22 009. 11

    work page 2021

  38. [38]

    Dropedge: Towards deep graph convolutional networks on node classification,

    Y . Rong, W. Huang, T. Xu, and J. Huang, “Dropedge: Towards deep graph convolutional networks on node classification,” in Proc. of ICLR. Addis Ababa, Ethiopia: OpenReview.net, 2020

    work page 2020

  39. [39]

    Graph augmentation clustering network,

    Z. Peng, H. Liu, Y . Jia, and J. Hou, “Graph augmentation clustering network,” 2022

    work page 2022

  40. [40]

    Position-aware structure learning for graph topology-imbalance by relieving under-reaching and over-squashing,

    Q. Sun, J. Li, H. Yuan, X. Fu, H. Peng, C. Ji, Q. Li, and P. S. Yu, “Position-aware structure learning for graph topology-imbalance by relieving under-reaching and over-squashing,” in Proc. of CIKM . Atlanta, GA, USA: ACM, 2022, pp. 1848–1857

    work page 2022

  41. [41]

    Deep fusion clustering network with reliable structure preservation,

    L. Gong, W. Tu, S. Zhou, L. Zhao, Z. Liu, and X. Liu, “Deep fusion clustering network with reliable structure preservation,” IEEE Transactions on Neural Networks and Learning Systems, pp. 1–12, 2022

    work page 2022

  42. [42]

    Visualizing data using t-sne,

    L. Van der Maaten and G. Hinton, “Visualizing data using t-sne,”Journal of Machine Learning Research , vol. 9, no. 11, pp. 2579–2605, 2008

    work page 2008

  43. [43]

    Geom-gcn: Geometric graph convolutional networks,

    H. Pei, B. Wei, K. C.-C. Chang, Y . Lei, and B. Yang, “Geom-gcn: Geometric graph convolutional networks,” in International Conference on Learning Representations , 2020

    work page 2020

  44. [44]

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

    J. Zhu, Y . Yan, L. Zhao, M. Heimann, L. Akoglu, and D. Koutra, “Beyond homophily in graph neural networks: Current limitations and effective designs,” Advances in neural information processing systems , vol. 33, pp. 7793–7804, 2020

    work page 2020

  45. [45]

    The generalization of ‘student’s’ problem when several different population variances are involved,

    B. L. Welch, “The generalization of ‘student’s’ problem when several different population variances are involved,” Biometrika, vol. 34, no. 1-2, pp. 28–35, 1947. Bneyu Wu received his B.Sc. degree from China University of Mining and Technology in 2022. He is currently pursuing the M.Sc. degree in China University of Mining and Technology. His research int...

    work page 1947

  46. [46]

    Her research interests include deep learning, clustering analysis, et al

    She is a lecturer at China University of Mining and Technology. Her research interests include deep learning, clustering analysis, et al. Lili Guo (Member, IEEE) received her Ph.D. degree from Tianjin University, Tianjin China in 2021. She is a lecturer at China University of Mining and Technology. Her research interests include deep learning, multimodal ...

    work page 2021