Pith. sign in

REVIEW 4 major objections 6 minor 39 references

GNN-MultiFix: Addressing the pitfalls for GNNs for multi-label node classification

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read GNN-MultiFix claims that multi-label node classification needs features, labels, and positions together, not just neighborhood aggregation, and that the combination makes any base GNN strictly more expressive.

desk verdict Useful modular method and a real empirical observation, but the 'improves all datasets' claim is contradicted by the paper's own tables and the error reporting is impossible; needs revision before acceptance. read the letter →

arxiv 2411.14094 v1 pith:4KM22AKT submitted 2024-11-21 cs.LG

classification cs.LG
keywords multi-labelnodeclassificationgraphneuralnetworkslabelpropagationpositionalencodingtransductivelearningisomorphismexpressivityDeepWalkembeddings
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Graph neural networks are usually judged on multi-class benchmarks or isomorphism tests, but the paper argues that this misses the harder, realistic case of multi-label node classification, where each node can carry several labels. It shows experimentally that standard GNNs underfit multi-label data even with abundant labeled nodes, and that a simple majority vote over labeled neighbors often beats them. The paper then claims that even the most expressive GNN, in the graph-isomorphism sense, cannot separate two nodes with identical local structure but different labels when node attributes are absent and labels are not fed in. To fix this, it proposes GNN-MultiFix, which combines three independent node representations—features, propagated labels, and a positional embedding—and proves that the combination is strictly more expressive than the base GNN. On four real-world and two synthetic multi-label datasets, every version of the method improves over all baselines.

What carries the argument

The load-bearing object is the three-module representation $\mathbf{h}_v = \mathrm{COMB}(\mathbf{h}^{(K)}_{f_v}, \mathbf{h}^{(N)}_{\ell_v}, \Phi_v)$: a feature module that aggregates one-hop neighborhoods with a GCN-style rule; a label module that initializes training nodes with true label vectors and test nodes with padding, then propagates for $N$ layers using the same weighted aggregation without the reset step that classical label propagation uses; and a positional module that embeds nodes by DeepWalk's random-walk co-occurrence, whose loss $\mathcal{L}(u,v) = (1 - I(u,v))\, f(\Phi_u, \Phi_v)$ pushes far-apart nodes apart. The proof that GNN-MultiFix strictly exceeds the base GNN's expressivity runs through Lemma 1: with a uniform random-walk transition matrix $P$, $N$ label-propagation steps give $\mathbf{h}_{\ell_v} = \sum_w P^N_{vw} \mathbf{y}_w$, so the label module encodes the label distribution of each node's $N$-hop training neighborhood; isomorphic nodes with different such distributions, or with different positions when the neighborhoods are unlabeled, become distinguishable.

What would settle it

Construct or find a multi-label graph family in which label assignment is independent of node positions—for example, two identically structured clusters with identical local neighborhoods but labels assigned independently of position—then compare GNN-MultiFix with and without the positional module. If removing the positional module does not change average precision, the claim that positional encoding is the mechanism behind the reported gains would be falsified.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes two failures: GNNs do not learn reliably from multi-label graphs—training dynamics show many atypical nodes whose losses stay high even when the mean training loss converges—and expressive power measured by graph isomorphism does not translate into distinguishing power for node labels, because two nodes with isomorphic computational graphs can sit in different parts of the network and have different labels. The proposed solution is to stop treating node labels as supervision only and instead feed three signals into the classifier: a standard GNN feature representation, a label representation obtained by propagating known labels through the graph without resetting training nodes to their true labels, and a positional representation from short random walks. Theorem 1 states that every node the base GNN can separate, GNN-MultiFix can also separate, and some nodes the base GNN conflates are separated by the label or position module. The abstract's empirical claim is that this yields significant improvement across all multi-label datasets tested.

Load-bearing premise

The method leans on the assumption that how close two nodes are in the graph, measured by short random walks, tracks how similar their labels are; if that fails for a graph family, the positional module that drives the reported gains stops working.

Editorial extensions

If this is right

  • If the central claim holds, a simple majority-vote baseline should become a standard comparison for any future GNN work on multi-label node classification, since it outperforms most existing GNNs on several real datasets.
  • GNN-MultiFix can be built on top of any base GNN as its feature module, so the expressivity gain transfers; the method is a plug-in rather than a replacement architecture.
  • The label propagation module's number of layers $N$ controls how far label influence reaches, which matters when labels are shared locally despite low overall homophily.
  • The positional module's random-walk co-occurrence provides a position signal complementary to both features and labels, and performance persists even when node features are entirely absent, as on BlogCat.
  • On the synthetic homophily sweep, the method nearly matches the majority-vote ceiling at high homophily while beating all other baselines, suggesting the gains are not just from exploiting trivial local label agreement.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A consequence the paper leaves implicit is that on featureless, low-homophily graphs the positional module carries almost the entire task; on any graph family where short random walks do not track label similarity, that module would likely stop helping and could hurt.
  • The strict expressivity proof holds for the general framework, but the implemented positional module uses learned DeepWalk embeddings rather than a deterministic position encoding, so the theorem's guarantee does not automatically apply to the exact configuration evaluated; swapping in a deterministic distance-based encoding would make the theory match the experiment.
  • The synthetic results where MajorityVote nearly solves the task suggest that some existing multi-label benchmarks are dominated by local label structure, so gains on those datasets should be read as gains on the hard core beyond what neighbor labels already reveal.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper studies multi-label node classification in the transductive setting. It argues that standard GNNs underperform simple label-only baselines such as MajorityVote, and it analyzes per-node training losses on BlogCat and DBLP to support the claim that GNNs fail to learn on multi-label data. The main proposal, GNN-MultiFix, combines three input modules: feature propagation, label propagation, and positional (DeepWalk) embeddings, and the paper proves a theoretical expressiveness result relative to the base GNN. Experiments on four real-world datasets and two synthetic families compare three GNN-MultiFix variants against twelve baselines and include an ablation study.

Significance. If substantiated, the paper would fill a real gap: multi-label node classification is understudied relative to multi-class classification, and the observation that simple MajorityVote beats many GNN designs is practically relevant. The proposed three-module architecture is simple and the code release is a positive contribution. However, the headline claim of improvement across all datasets is contradicted by the paper's own reported numbers, and the reported standard deviations are internally inconsistent, so the statistical significance claim cannot currently be assessed. The theoretical expressiveness result is also not matched by the implemented positional module. The central idea is defensible, but the current evidence requires substantial qualification.

major comments (4)
  1. [Abstract, Section 6, Conclusion; Tables 3 and 4] The abstract and conclusion claim that GNN-MultiFix achieves significant improvement across all multi-label datasets, but the paper's own results contradict this. In Table 3, on Yelp, GraphSage achieves Average Precision 0.251 while the best GNN-MultiFix variant (MLP3) achieves 0.238. In Table 4, on Synthetic2 with label homophily 1.0, MajorityVote achieves 0.847 while the best GNN-MultiFix variant (MLP1) achieves 0.832, and at homophily 0.2, Mlp achieves 0.362 while the best GNN-MultiFix variant achieves 0.288. The claim should be qualified to 'most settings' and the text in Section 6.2 ('our model and its variants demonstrate significant enhancements') should be aligned with these counterexamples.
  2. [Appendix A.2.4, Tables 9-11] The reported standard deviations are incompatible with Average Precision being bounded in [0,1]. For example, Table 9 reports Mlp on BlogCat as 0.043 ± 0.63, DeepWalk on PCG as 0.229 ± 1.00, and GCN-LPA on DBLP as 0.801 ± 1.24; Table 10 reports DeepWalk as 0.181 ± 0.71; Table 11 reports DeepWalk as 0.813 ± 0.95. Since the word 'significant' is load-bearing, the authors must either correct the variability measures or provide proper significance tests. As written, the statistical evidence cannot be evaluated.
  3. [Section 4.2, Lemma 1 and Theorem 1; Section 4.1.3] The theoretical expressiveness claim is only partially supported by the implemented method. Lemma 1 shows that, by feeding true training labels as input (Eq. 2), the label representation becomes a weighted sum of neighbor labels; this is an expressiveness gain by construction, not a property of the GNN architecture itself. Theorem 1 then claims that isomorphic test nodes can be distinguished using positional representations, but the implemented module uses DeepWalk embeddings, which are not guaranteed to assign different vectors to nodes with isomorphic computational graphs; in symmetric graphs two such nodes may have identical random-walk statistics. The proof should either state conditions under which the implemented Φv separates such nodes or restrict the claim to the theoretical instantiation (e.g., Katz similarity) and verify that the implementation satisfies those conditions.
  4. [Section 3 and Figures 3-4] The claim that GNNs 'fail to learn' on multi-label data is based on visual inspection of box plots of per-node training losses without quantitative thresholds, convergence criteria, or comparison across seeds. As presented, this is not a falsifiable empirical analysis. The authors should add quantitative evidence, such as the fraction of nodes whose loss remains above a defined threshold at convergence, to support the failure-to-learn claim.
minor comments (6)
  1. [Section 5, Baselines list] The baseline enumeration uses 'iiii' for the fourth group; the numbering should be corrected to a consistent scheme.
  2. [Table 7 caption] The caption contains the typo 'numebr' instead of 'number'.
  3. [Tables 3 and 9] Table 3 and Table 9 report slightly different values for the same method on Yelp, e.g., GNN-MultiFix-Linear 0.200 vs 0.201 and GNN-MultiFix-MLP3 0.238 vs 0.237; the authors should ensure the two tables are fully consistent or explain the source of the discrepancy.
  4. [Section 4.1.3, Eq. (4)] The positional encoding loss is described only abstractly via the functions I and f; the text should specify the concrete objective used in the implementation, for example the exact SkipGram negative-sampling formulation, so that the method is reproducible from the paper alone.
  5. [Section 7, Table 6] On BlogCat, the full GNN-MultiFix-Linear achieves exactly the same score (0.225) as the variant without the feature module; the text says removing the feature module has 'little effect,' but the exact equality deserves explicit discussion.
  6. [Section 4.1.2, Eq. (2)] The use of true training labels as input to the label representation module should be discussed explicitly as a potential source of train/test distribution shift, especially for datasets with very few labels per node, since this choice is central to the method's design.

Circularity Check

1 steps flagged · score 4.0 of 10

The theoretical expressiveness theorem is self-definitional because it feeds true labels as input by construction, but the empirical benchmark comparison is independent; the score reflects only partial circularity.

  1. self definitional [Section 4.2, Lemma 1 and Theorem 1 (Eqs. 2-3)]
    "Executing the aggregation function N times result in the label representation of the node v as h_lv = ∑_{w∈V} P^N_{vw} y_w ... In the simpler case, the label distribution of the N-hop neighborhood of the isomorphic nodes is known and is different. In this case by Lemma 1 the isomorphic nodes can be distinguished by their label representations."

    The theorem's first distinguishing case is true by construction: the label representation module is defined, via Eq. (2)-(3), as a weighted sum of the input true label vectors y_w. Whenever two nodes have different known N-hop label distributions, their label representations differ because the inputs differ, independent of any learned parameters or generalization. The claimed expressiveness gain over the base GNN therefore reduces to 'we fed different label vectors as input to those nodes.' The second case defers to a hypothetical Katz-style positional representation rather than the implemented DeepWalk embeddings, so no additional nontrivial derivation is supplied.

full rationale

The empirical evaluation is largely self-contained: GNN-MultiFix is compared against 12 baselines on real and synthetic datasets, so the headline performance claim does not reduce to the paper's own fitted parameters. At the same time, the theoretical expressiveness argument in Section 4.2 is self-definitional. Lemma 1 instantiates the label module as a random-walk weighted sum of the true input label vectors, and Theorem 1's first distinguishing case then follows simply because different input label distributions yield different weighted sums. This is a restatement of the input construction rather than a derived property. The second case is not implemented (it invokes Katz similarity via [13], while experiments use DeepWalk), so the 'strictly more expressive' theorem is not independently established. Because the empirical results do not depend on this theorem, the circularity is partial. Separately, the abstract's 'all multi-label datasets' claim is contradicted by Tables 3 and 4 (Yelp: best GNN-MultiFix 0.238 vs GraphSage 0.251; Synthetic2 homophily=1.0: best GNN-MultiFix 0.832 vs MajorityVote 0.847), and some reported standard deviations are impossible for AP in [0,1]; these are correctness concerns, not circularity, and are noted here for completeness.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central method rests on two domain assumptions in Section 4.1.3 about distance-limited influence and path-count label similarity, on the availability of complete training labels, and on the existence of positional embeddings that can distinguish non-isomorphic nodes. Hyperparameters such as label propagation depth N and embedding dimension are tuned per dataset. No new physical or conceptual entities are introduced.

free parameters (3)
  • label propagation depth N = BlogCat 1, Yelp 25, DBLP 2, PCG 1 or 5
    Tuned per dataset in Table 8; controls the receptive field of the label module and directly shapes the label representation.
  • DeepWalk embedding dimension = 64
    Chosen by hand for the positional module; the paper does not report sensitivity to this choice.
  • walk length and window size for DeepWalk = length 10, window 5
    Fixed implementation choices for positional embeddings, not justified by analysis.
assumptions (4)
  • domain assumption Distance limitation on influence: nodes far apart are less likely to influence each other, and k-hop neighborhood suffices to reconstruct labels for small k.
    Stated in Section 4.1.3 and supported only by 'preliminary analysis' that is not shown; the ablation on BlogCat shows the positional module is critical, so this assumption is load-bearing.
  • domain assumption Influence-driven label similarity: label similarity between two nodes increases with the number of paths for influence propagation between them.
    Also in Section 4.1.3; used to justify learning positional embeddings with random walks.
  • standard math Definition of 'most expressive GNN' (Definition 3): a GNN that maps isomorphic graphs to the same representation and distinguishes non-isomorphic graphs.
    Inherited from the Weisfeiler-Lehman expressiveness literature; the paper does not prove such a GNN exists for all cases.
  • domain assumption Training nodes are completely labeled.
    Stated in the Problem Setting section; the label module uses these full labels as input.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GNN-MultiFix: Addressing the pitfalls for GNNs for multi-label node classification." pith.science (2026). https://pith.science/paper/4KM22AKT

@misc{pith2026241114094,
  author       = {Pith},
  title        = {Pith review of: GNN-MultiFix: Addressing the pitfalls for GNNs for multi-label node classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4KM22AKT}},
  note         = {Machine review of arXiv:2411.14094}
}
read the original abstract

Graph neural networks (GNNs) have emerged as powerful models for learning representations of graph data showing state of the art results in various tasks. Nevertheless, the superiority of these methods is usually supported by either evaluating their performance on small subset of benchmark datasets or by reasoning about their expressive power in terms of certain graph isomorphism tests. In this paper we critically analyse both these aspects through a transductive setting for the task of node classification. First, we delve deeper into the case of multi-label node classification which offers a more realistic scenario and has been ignored in most of the related works. Through analysing the training dynamics for GNN methods we highlight the failure of GNNs to learn over multi-label graph datasets even for the case of abundant training data. Second, we show that specifically for transductive node classification, even the most expressive GNN may fail to learn in absence of node attributes and without using explicit label information as input. To overcome this deficit, we propose a straightforward approach, referred to as GNN-MultiFix, that integrates the feature, label, and positional information of a node. GNN-MultiFix demonstrates significant improvement across all the multi-label datasets. We release our code at https://anonymous.4open.science/r/Graph-MultiFix-4121.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 24 canonical work pages

  1. [1]

    Estimating example difficulty using variance of gradients

    Chirag Agarwal and Sara Hooker. Estimating example difficulty using variance of gradients. CoRR, abs/2008.11600, 2020. URL https://arxiv.org/abs/2008.11600

  2. [2]

    Collaborative Graph Walk for Semi-supervised Multi-Label Node Classification

    Uchenna Akujuobi, Yufei Han, Qiannan Zhang, and Xiangliang Zhang. Collab- orative graph walk for semi-supervised multi-label node classification.CoRR, abs/1910.09706, 2019. URL http://arxiv.org/abs/1910.09706

  3. [3]

    THE LOGICAL EXPRESSIVENESS OF GRAPH NEURAL NETWORKS

    Pablo Barceló, Egor V Kostylev, Mikaël Monet, Jorge Pérez, Juan Reutter, and Juan-Pablo Silva. THE LOGICAL EXPRESSIVENESS OF GRAPH NEURAL NETWORKS. In 8th International Conference on Learning Representations (ICLR 2020), Virtual conference, Ethiopia, April 2020. URL https://hal.science/ hal-03356968

  4. [4]

    Multi-label image recognition with graph convolutional networks

    Zhao-Min Chen, Xiu-Shen Wei, Peng Wang, and Yanwen Guo. Multi-label image recognition with graph convolutional networks. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5177–5186, 2019

  5. [5]

    Towards a consistent evaluation of mirna- disease association prediction models

    Thi Ngan Dong and Megha Khosla. Towards a consistent evaluation of mirna- disease association prediction models. In2020 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pages 1835–1842. IEEE, 2020

  6. [6]

    Graph neural networks with learnable structural and positional representations

    Vijay Prakash Dwivedi, Anh Tuan Luu, Thomas Laurent, Yoshua Bengio, and Xavier Bresson. Graph neural networks with learnable structural and positional representations. InInternational Conference on Learning Representations, 2022. URL https://openreview.net/forum?id=wTTjnvGphYj

  7. [7]

    Hamilton, Rex Ying, and Jure Leskovec

    William L. Hamilton, Rex Ying, and Jure Leskovec. Inductive representation learning on large graphs.CoRR, abs/1706.02216, 2017. URL http://arxiv.org/ abs/1706.02216

  8. [8]

    Prentice Hall PTR, 1994

    Simon Haykin.Neural networks: a comprehensive foundation. Prentice Hall PTR, 1994

Show all 39 references
  1. [9]

    Multi-label learning by exploiting label correlations locally.Proceedings of the AAAI Conference on Artificial Intelligence, 26(1):949–955, Sep

    Sheng-Jun Huang and Zhi-Hua Zhou. Multi-label learning by exploiting label correlations locally.Proceedings of the AAAI Conference on Artificial Intelligence, 26(1):949–955, Sep. 2021. doi: 10.1609/aaai.v26i1.8287. URL https://ojs.aaai. org/index.php/AAAI/article/view/8287

  2. [10]

    Semi-supervised classification with graph convolutional networks.arXiv preprint arXiv:1609.02907, 2016

    Thomas N Kipf and Max Welling. Semi-supervised classification with graph convolutional networks.arXiv preprint arXiv:1609.02907, 2016

  3. [11]

    Neural message passing for multi-label classification

    Jack Lanchantin, Arshdeep Sekhon, and Yanjun Qi. Neural message passing for multi-label classification. In Joint European Conference on Machine Learning 17 and Knowledge Discovery in Databases, pages 138–163. Springer, 2019

  4. [12]

    Improving graph neural networks with simple architecture design, 2021

    Sunil Kumar Maurya, Xin Liu, and Tsuyoshi Murata. Improving graph neural networks with simple architecture design, 2021. URL https://arxiv.org/abs/2105. 07634

  5. [13]

    Asymmetric transitivity preserving graph embedding

    Mingdong Ou, Peng Cui, Jian Pei, et al. Asymmetric transitivity preserving graph embedding. In Proc. of the International Conference on Knowledge Discovery and Data Mining, pages 1105–1114, 2016

  6. [14]

    Deepwalk: Online learning of social representations

    Bryan Perozzi, Rami Al-Rfou, and Steven Skiena. Deepwalk: Online learning of social representations. InProc. of the International Conference on Knowledge Discovery and Data Mining, 2014

  7. [15]

    Galaxc: Graph neural networks with labelwise attention for extreme classification

    Deepak Saini, Arnav Kumar Jain, Kushal Dave, Jian Jiao, Amit Singh, Ruofei Zhang, and Manik Varma. Galaxc: Graph neural networks with labelwise attention for extreme classification. WWW ’21, New York, NY, USA, 2021. Association for Computing Machinery. ISBN 9781450383127. doi:...

  8. [16]

    Training-free graph neural networks and the power of labels as features, 2024

    Ryoma Sato. Training-free graph neural networks and the power of labels as features, 2024

  9. [17]

    Multi-label graph con- volutional network representation learning.CoRR, abs/1912.11757, 2019

    Min Shi, Yufei Tang, Xingquan Zhu, and Jianxun Liu. Multi-label graph con- volutional network representation learning.CoRR, abs/1912.11757, 2019. URL http://arxiv.org/abs/1912.11757

  10. [18]

    Metadata archaeology: Unearthing data subsets by leveraging training dynamics, 2022

    Shoaib Ahmed Siddiqui, Nitarshan Rajkumar, Tegan Maharaj, David Krueger, and Sara Hooker. Metadata archaeology: Unearthing data subsets by leveraging training dynamics, 2022

  11. [19]

    Semi-supervised multi- label learning for graph-structured data

    Zixing Song, Ziqiao Meng, Yifei Zhang, and Irwin King. Semi-supervised multi- label learning for graph-structured data. In Proceedings of the 30th ACM International Conference on Information & Knowledge Management, pages 1723–1733, 2021

  12. [21]

    On the equivalence between positional node embeddings and structural graph representations.arXiv preprint arXiv:1910.00452, 2019

    Balasubramaniam Srinivasan and Bruno Ribeiro. On the equivalence between positional node embeddings and structural graph representations.arXiv preprint arXiv:1910.00452, 2019

  13. [22]

    Smith, and Yejin Choi

    Swabha Swayamdipta, Roy Schwartz, Nicholas Lourie, Yizhong Wang, Hannaneh Hajishirzi, Noah A. Smith, and Yejin Choi. Dataset cartography: Mapping and diagnosing datasets with training dynamics.CoRR, abs/2009.10795, 2020. URL https://arxiv.org/abs/2009.10795

  14. [23]

    Relational learning via latent social dimensions

    Lei Tang and Huan Liu. Relational learning via latent social dimensions. In Proceedings of the 15th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’09, page 817–826, New York, NY, USA,

  15. [24]

    Graph Attention Networks.International Conference on Learning Representations, 2018

    Petar Veličković, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Liò, and Yoshua Bengio. Graph Attention Networks.International Conference on Learning Representations, 2018. URL https://openreview.net/forum?id= 18 rJXMpikCZ. accepted as poster

  16. [25]

    Bo Wang, Zhuowen Tu, and John K. Tsotsos. Dynamic label propagation for semi-supervised multi-class multi-label classification. In2013 IEEE International Conference on Computer Vision, pages 425–432, 2013. doi: 10.1109/ICCV.2013.60

  17. [26]

    Unifying graph convolutional neural networks and label propagation.CoRR, abs/2002.06755, 2020

    Hongwei Wang and Jure Leskovec. Unifying graph convolutional neural networks and label propagation.CoRR, abs/2002.06755, 2020. URL https://arxiv.org/abs/ 2002.06755

  18. [27]

    How powerful are graph neural networks?CoRR, abs/1810.00826, 2018

    Keyulu Xu, Weihua Hu, Jure Leskovec, and Stefanie Jegelka. How powerful are graph neural networks?CoRR, abs/1810.00826, 2018. URL http://arxiv.org/abs/ 1810.00826

  19. [28]

    Extract the knowledge of graph neural networks and go beyond it: An effective knowledge distillation framework

    Cheng Yang, Jiawei Liu, and Chuan Shi. Extract the knowledge of graph neural networks and go beyond it: An effective knowledge distillation framework. In Proceedings of The Web Conference 2021 (WWW ’21). ACM, 2021

  20. [29]

    Breaking the expression bottleneck of graph neural networks

    Mingqi Yang, Renjian Wang, Yanming Shen, Heng Qi, and Baocai Yin. Breaking the expression bottleneck of graph neural networks. IEEE Transactions on Knowledge and Data Engineering, 35(6):5652–5664, 2023. doi: 10.1109/TKDE. 2022.3168070

  21. [30]

    Evaluating link prediction methods

    Yang Yang, Ryan N Lichtenwalter, and Nitesh V Chawla. Evaluating link prediction methods. Knowledge and Information Systems, 45(3):751–782, 2015

  22. [31]

    Identity- aware graph neural networks.CoRR, abs/2101.10320, 2021

    Jiaxuan You, Jonathan Gomes-Selman, Rex Ying, and Jure Leskovec. Identity- aware graph neural networks.CoRR, abs/2101.10320, 2021. URL https://arxiv. org/abs/2101.10320

  23. [32]

    Prasanna

    Hanqing Zeng, Hongkuan Zhou, Ajitesh Srivastava, Rajgopal Kannan, and Vik- tor K. Prasanna. Graphsaint: Graph sampling based inductive learning method. CoRR, abs/1907.04931, 2019. URL http://arxiv.org/abs/1907.04931

  24. [33]

    Eigen-gnn: A graph structure preserving plug-in for gnns.CoRR, abs/2006.04330, 2020

    Ziwei Zhang, Peng Cui, Jian Pei, Xin Wang, and Wenwu Zhu. Eigen-gnn: A graph structure preserving plug-in for gnns.CoRR, abs/2006.04330, 2020. URL https://arxiv.org/abs/2006.04330

  25. [34]

    Multi-label node classification on graph-structured data.Transactions on Machine Learn- ing Research, 2023

    Tianqi Zhao, Thi Ngan Dong, Alan Hanjalic, and Megha Khosla. Multi-label node classification on graph-structured data.Transactions on Machine Learn- ing Research, 2023. ISSN 2835-8856. URL https://openreview.net/forum?id= EZhkV2BjDP

  26. [35]

    Towards data-centric graph machine learning: Review and outlook

    Xin Zheng, Yixin Liu, Zhifeng Bao, Meng Fang, Xia Hu, Alan Wee-Chung Liew, and Shirui Pan. Towards data-centric graph machine learning: Review and outlook. arXiv preprint arXiv:2309.10979, 2023

  27. [36]

    Cangqi Zhou, Hui Chen, Jing Zhang, Qianmu Li, Dianming Hu, and Victor S. Sheng. Multi-label graph node classification with label attentive neighborhood convolution. Expert Systems with Applications, 180:115063, 2021. ISSN 0957-4174. doi: https://doi.org/10.1016/j.eswa.2021.115...

  28. [37]

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

    Jiong Zhu, Yujun Yan, Lingxiao Zhao, Mark Heimann, Leman Akoglu, and Danai Koutra. Beyond homophily in graph neural networks: Current limitations and effective designs. Advances in Neural Information Processing Systems, 33, 2020

  29. [38]

    N.A." and

    Yue Zhu, James T. Kwok, and Zhi-Hua Zhou. Multi-label learning with global and local label correlation.CoRR, abs/1704.01415, 2017. 19 A Appendix Organization of the Appendix.In Section A.1, we provide pseudo code for the algorithm of GNN-MultiFix. In the following Section A.2,...

  30. [41]

    OOM" denotes the

    We use the consistant notations as in the main paper.k represents the number of feature propagation,N indicates the number of label propagation. We also used early stopping with the patience specified in the table to prevent the models from overfitting. 21 The hidden dimension...

  31. [2009]

    ISBN 9781605584959

    Association for Computing Machinery. ISBN 9781605584959. doi: 10.1145/ 1557019.1557109. URL https://doi.org/10.1145/1557019.1557109

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.