Pith. sign in

REVIEW 4 major objections 6 minor 48 references

Training a Label-Noise-Resistant GNN with Reduced Complexity

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

Pith's one-line read This paper claims that graph node classification under label noise can be made both more accurate and cheaper by ensembling many candidate labels from randomly masked neighbor views instead of certifying one reliable label per node.

desk verdict A well-executed empirical method paper worth refereeing, with a misprinted core equation and an untested homophily assumption. read the letter →

arxiv 2411.11020 v1 pith:FRFUIDSN submitted 2024-11-17 cs.LG cs.SI

classification cs.LGcs.SI
keywords graphneuralnetworkslabelnoisenodeclassificationpartiallearningensemblesemi-supervisednoisylabelsscalability
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

LEGNN claims that noisy-label node classification on graphs is better framed as collecting a set of plausible labels per node than as deciding one reliable label. Instead of computing pairwise node similarities to filter noise, it randomly masks each node's neighbors several times, lets the current GNN label each masked view, and unions the top predictions into a high-probability label set plus a symmetric low-probability set. A partial-label-learning weighted loss trains the GNN from these sets, so a single mistaken pseudo-label cannot by itself steer training. The paper reports that this label-ensemble view beats reliable-labeling methods such as NRGNN and RTGNN on six datasets, with the largest gains at high noise rates, while adding only near-linear-to-subquadratic cost rather than $O(N^2)$.

What carries the argument

The load-bearing object is the bootstrapped neighbor-context label ensemble, built on top of partial label learning, a setting in which each training example is given a set of candidate labels known to contain the true label. A random edge mask with rate $K$ is applied $M_e$ times to create diverse graph views; the current GNN labels each view, and the ensemble keeps every label that ever appears as an arg-max (high-probability set) and every label that ever appears as an arg-min (low-probability set). The second piece is the weighted bidirectional loss, which trains on both sets at once and makes the method, in the authors' account, shift from precision-oriented single-label correction to a precision-recall-balanced ensemble.

What would settle it

On a graph whose edges predominantly connect nodes of different classes (high heterophily), directly measure the fraction of nodes whose true label actually appears in the high-probability label set under 50% symmetric label noise; if that fraction is near chance while accuracy still improves, the homophily mechanism is not the source of the gain.

Watch

Extended reading notes

Core claim

Central claim: the reason previous noise-robust GNNs degrade at high noise is that they sacrifice recall for precision when they assign one reliable pseudo-label, and LEGNN can have both by ensembling labels across bootstrapped neighbor contexts. In LEGNN, the model first predicts each node under $M_e$ random edge masks; for each node the arg-max predictions form the high-probability multi-label set and the arg-min predictions form the low-probability multi-label set. Training then minimizes a weighted bidirectional loss, $L(Z,Y^p,Y^n)=L_p(Z,Y^p)+L_p(O-Z,Y^n)$, where the partial-label-learning weighting down-weights low-confidence labels and the low-probability term supplies negative supervision. The paper shows on Cora, Citeseer, Pubmed, IGB, OGBN-Arxiv, and a Clothing1m KNN graph that this yields accuracy gains that grow with noise, e.g. 14.55 points on Citeseer and 8.47 points on OGBN-Arxiv at 50% symmetric noise, and that the additional complexity is $O(M_e(1+Ld)|E|+(M_eLd^2+2C)N)$, avoiding the $O(N^2)$ of similarity-based reliable labeling.

Load-bearing premise

LEGNN's advantage depends on homophily: a node's true label must usually be among the labels of its neighbors, so that the randomly masked neighbor views and the model's bootstrapped predictions keep the correct label inside the collected high-probability set.

Editorial extensions

If this is right

  • Noisy-label node classification can be handled without expensive pairwise-similarity reliability modules, so robustness scales to graphs with hundreds of thousands of nodes and millions of edges.
  • Performance degrades more slowly as noise increases; the reported margin over NRGNN and RTGNN is largest at 50% symmetric and 40% pair-flip noise.
  • The same training wrapper works with GCN, GAT, and GraphSAGE backbones, so it can be dropped onto an existing GNN rather than requiring a bespoke architecture.
  • Gathered label sets cover more true labels than confidence-threshold selection, which explains why training on noisier candidate sets can still improve accuracy.
  • The method's additional cost stays below $O(N^2)$ except on nearly complete graphs, since it depends on edge count and label-ensemble size rather than all-pairs similarity.

Reading between the lines

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

  • A natural stress test beyond the paper is heterophily: because the bootstrapping argument assumes neighbors are informative about the true label, LEGNN should be evaluated on graphs where edges mostly connect different classes; the homophily premise predicts a sharp drop.
  • The random-mask ensemble is essentially a cheap way to create diversity; replacing the fixed arg-max/arg-min rule with sampled labels or temperature-scaled probabilities could interpolate between precision and recall and possibly improve the reported trade-off.
  • The label-gathering step is only triggered when validation accuracy drops, so the amortized cost is lower than the worst-case complexity suggests; a systematic study of gathering frequency could sharpen the efficiency claim.
  • If the precision-recall diagnosis is correct, other pseudo-label methods could adopt the same symmetric low-probability negative supervision without any graph-specific machinery, transferring the idea to image or text noisy-label learning.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 proposes LEGNN, a label-noise-resistant GNN training method that reframes noisy-label node classification as a partial label learning (PLL) task. Instead of constructing a single reliable label, LEGNN generates, for each node, a high-probability label set and a low-probability label set from predictions on several randomly edge-masked versions of the graph, and trains the GNN with a weighted bidirectional loss. The authors claim that this label-ensemble approach avoids the O(N^2) cost of similarity-based reliability assessment while achieving state-of-the-art accuracy on noisy node classification benchmarks. Experiments are reported on Cora, Citeseer, Pubmed, IGB, OGBN-Arxiv, and a KNN-constructed Clothing1m graph, under symmetric and pair noise, plus an analysis of label precision/recall and runtime.

Significance. If the claims hold, LEGNN is a meaningful contribution: it offers an alternative to reliable-labeling methods, is conceptually simple, and the reported speedups over NRGNN/RTGNN are substantial (e.g., Table III). The paper also provides a useful empirical analysis showing that higher recall of true labels can compensate for lower pseudo-label precision (Table VI and VII). The code is publicly available, and the experiments cover large-scale graphs (IGB, OGBN-Arxiv), which strengthens the scalability claim. The main weakness is that the theoretical motivation in Section IV-A is a toy model that does not match the actual algorithm, and the evaluation is restricted to homophilous graphs, so the generality of the 'superior performance' claim is not yet established.

major comments (4)
  1. [Eq. (11)] The weighting function in Eq. (11) is printed as w(Zi, Yp_ij) = Zij Yp_ij / sum_k Zik Yp_ij. The denominator is independent of the summation index k, and when Yp_ij = 0 the denominator is zero, making the weight undefined. This is presumably a typo for sum_k Zik Yp_ik, which is the standard PLL weighting (e.g., Feng et al. [24]). Because this equation defines the core loss, it must be corrected and the surrounding text updated to match.
  2. [Sec. IV-A] The theoretical justification for bootstrapped neighbor masking is a majority-vote bound that assumes each neighbor's label is erroneous independently with probability alpha <= 0.5. LEGNN does not perform such a vote: it takes argmax/argmin predictions from the same GNN on Me masked graphs, and those predictions are strongly correlated because they share the same model and largely overlapping neighborhoods. The independence assumption is therefore not satisfied, and the cited argument does not establish that the gathered high-probability set contains the true label. The paper should either provide an analysis that matches the actual algorithm or clearly label the discussion as an intuition only, without claiming a noise-reduction guarantee.
  3. [Sec. IV-A and Sec. V] LEGNN explicitly relies on the homophily assumption stated at the start of Section IV-A. Under heterophily, or when the constructed KNN graph is uninformative, the true label can be systematically absent from the high-probability set and present in the low-probability set; Eq. (12) would then actively penalize the correct class. All six benchmarks are homophilous citation networks or a KNN graph built from ImageNet features, so this failure mode is untested. The conclusion that LEGNN achieves 'superior performance' should be qualified to the homophilous setting, or the authors should add experiments on heterophilous benchmarks (e.g., Chameleon, Squirrel, Actor) and discuss how the method behaves when the assumption is violated.
  4. [Sec. IV-E] The claimed additional complexity O(Me(1+Ld)|E| + (MeLd^2+2C)N) omits the cost of computing argmax and argmin over C classes for every node and every masked graph, which is O(Me N C). This does not change the asymptotic advantage over O(N^2) when C is a constant, but the stated formula is incomplete and should be corrected. Please also clarify whether the 2CN term accounts for the loss computation only; as written, the argmax/argmin cost is missing.
minor comments (6)
  1. [Eqs. (7) and (9)] The notation for the gathered label sets is inconsistent: Eq. (6) defines \hat y_k^i, while Eqs. (7) and (9) use \hat y_k^i and \bar y_k^i with the indices in different positions. Please unify the superscript/subscript convention throughout.
  2. [Sec. V-C] The subsection beginning 'Discuss: why NRGNN and RTGNN perform worse than GCN in Pubmed and OGBN?' is written in an informal Q&A style and interrupts the flow of the experimental section. It should be integrated into the main text or moved to an appendix with a proper framing.
  3. [Alg. 1] Algorithm 1, step 6, says 'if validation classification accuracy declined' without specifying the threshold, the frequency of checking, or how the noisy validation set is used. This condition is underspecified and should be described precisely, since it controls when the label gathering is repeated.
  4. [Fig. 1 caption] The caption states 'The experimental details of (b) are inline with Section V-E', but Section V-E does not contain a description of Figure 1(b). Please either add the referenced details or remove the cross-reference.
  5. [General] There are several typos and minor wording issues, e.g., 'exibit imitations' (Sec. II-A), 'diagnal matrix' (Eq. 3), and 'PPL' vs 'PLL' (Sec. III-C). These should be cleaned up.
  6. [References] Reference [43] is cited for the complexity of an L-layer GCN, but the title of that reference is about molecular property prediction; it is unclear that this is the intended source. Please cite a standard GCN complexity reference instead.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the central claim is an empirical benchmark comparison, and the only in-house citations are non-load-bearing application references.

full rationale

LEGNN's advertised result is test-accuracy superiority over baselines under injected and real label noise, plus a complexity bound. Tracing the derivation chain, the label-ensemble construction (Eqs. 6-9), the weighted bidirectional PLL loss (Eqs. 10-13), and the complexity analysis (Sec. IV-E) do not take the reported accuracies as inputs. The high- and low-probability label sets are forward-pass outputs of the current model on masked graphs; the loss updates the model; and the evaluation is on held-out clean labels. This is a self-training/bootstrap loop rather than a claim that a derived quantity equals a fitted quantity, so it does not reduce to its inputs by construction. The PLL weighting is imported from external work [24] (Feng et al.), not from the authors' own results. The only in-house citations ([5,6,15]) appear in motivational sentences about taxpayer and traffic graphs and are not load-bearing for the label-ensemble method. The homophily premise is cited to external literature [42,16], and the paper tests only homophilous benchmarks plus a KNN-constructed Clothing1m graph. The gap between the voting toy model (which assumes independent neighbor errors) and the actual algorithm (which aggregates correlated argmax/argmin predictions from the same model over masked graphs) is a validity concern, not a circular reduction. The manuscript's inserted 'Discuss:' paragraph is a post-hoc hypothesis for baseline failures and does not constitute circularity either. No equation in the paper is equivalent to the target result by definition, and no fitted parameter is renamed as a prediction. Score 2 reflects only the minor, non-load-bearing self-citations.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central method rests on a homophily assumption, on the conjecture that neighbor noise stays below 50%, on the imported consistency of the weighted PLL loss, and on the heuristic that low-probability labels provide useful negative supervision. The two main hyperparameters (mask rate and number of masks) are tuned on validation data. No new entities are introduced.

free parameters (2)
  • mask rate K = 0.4-0.6 optimal on Citeseer at high noise; searched over {0.1,...,0.9}
    Controls how many neighbors are masked per bootstrap; accuracy varies with K especially under high noise, and it is chosen via validation accuracy.
  • mask iterations Me = 10-20 optimal on Citeseer; searched over {5,...,30}
    Controls the number of bootstrapped graphs and therefore the diversity and size of the label ensemble; selected by validation accuracy.
assumptions (4)
  • domain assumption Neighboring nodes connected by edges often share the same label (homophily).
    Invoked in Section IV-A to justify using neighbor contexts for label bootstrap; the method's label ensembles are only informative if local neighborhoods carry label signal. Experiments are on homophilous or KNN-constructed graphs, so the assumption is untested on heterophilous graphs.
  • domain assumption The neighbor noise rate is less than 0.5, so majority-style bootstrap reduces error.
    Stated in Section IV-A discussion as a 'mild assumption'; the toy binomial argument requires per-neighbor error probability alpha <= 0.5 and independent errors.
  • domain assumption The loss-weighted PLL strategy of Feng et al. applies to the gathered high-probability candidate sets and the symmetric low-probability loss.
    Section IV-C imports the weighted PLL loss and assumes candidate sets contain the true label with sufficient probability; no theorem is given for the graph ensemble setting.
  • domain assumption Low-probability labels provide useful negative supervision that counteracts noise in the high-probability set.
    Section IV-C constructs Ln = Lp(O-Z, Yn) without proof that complement probabilities act as valid negative signals.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Training a Label-Noise-Resistant GNN with Reduced Complexity." pith.science (2026). https://pith.science/paper/FRFUIDSN

@misc{pith2026241111020,
  author       = {Pith},
  title        = {Pith review of: Training a Label-Noise-Resistant GNN with Reduced Complexity},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FRFUIDSN}},
  note         = {Machine review of arXiv:2411.11020}
}
read the original abstract

Graph Neural Networks (GNNs) have been widely employed for semi-supervised node classification tasks on graphs. However, the performance of GNNs is significantly affected by label noise, that is, a small amount of incorrectly labeled nodes can substantially misguide model training. Mainstream solutions define node classification with label noise (NCLN) as a reliable labeling task, often introducing node similarity with quadratic computational complexity to more accurately assess label reliability. To this end, in this paper, we introduce the Label Ensemble Graph Neural Network (LEGNN), a lower complexity method for robust GNNs training against label noise. LEGNN reframes NCLN as a label ensemble task, gathering informative multiple labels instead of constructing a single reliable label, avoiding high-complexity computations for reliability assessment. Specifically, LEGNN conducts a two-step process: bootstrapping neighboring contexts and robust learning with gathered multiple labels. In the former step, we apply random neighbor masks for each node and gather the predicted labels as a high-probability label set. This mitigates the impact of inaccurately labeled neighbors and diversifies the label set. In the latter step, we utilize a partial label learning based strategy to aggregate the high-probability label information for model training. Additionally, we symmetrically gather a low-probability label set to counteract potential noise from the bootstrapped high-probability label set. Extensive experiments on six datasets demonstrate that LEGNN achieves outstanding performance while ensuring efficiency. Moreover, it exhibits good scalability on dataset with over one hundred thousand nodes and one million edges.

Figures

Figures reproduced from arXiv: 2411.11020 by the authors.

Figure 1
Figure 1. A comparison between our proposed label ensemble [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The overall framework of LEGNN. The nodes are [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The classification accuracy decreases as the number of [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Classification accuracy (%) with different parameters [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 37 canonical work pages

  1. [24]

    Provably consistent partial- label learning,

    L. Feng, J. Lv, B. Han, M. Xu, G. Niu, X. Geng, B. An, and M. Sugiyama, “Provably consistent partial- label learning,” Advances in neural information process- ing systems, vol. 33, pp. 10 948–10 960, 2020

  2. [1]

    Adversarial label-flipping attack and defense for graph neural net- works,

    M. Zhang, L. Hu, C. Shi, and X. Wang, “Adversarial label-flipping attack and defense for graph neural net- works,” in 2020 IEEE International Conference on Data Mining (ICDM). IEEE, 2020, pp. 791–800

  3. [2]

    Inductive rep- resentation learning on large graphs,

    W. Hamilton, Z. Ying, and J. Leskovec, “Inductive rep- resentation learning on large graphs,” Advances in neural information processing systems , vol. 30, 2017

  4. [3]

    Modeling network- level traffic flow transitions on sparse data,

    X. Lei, H. Mei, B. Shi, and H. Wei, “Modeling network- level traffic flow transitions on sparse data,” in Proceed- ings of the 28th ACM SIGKDD Conference on Knowl- edge Discovery and Data Mining , 2022, pp. 835–845

  5. [4]

    Stochastic weight completion for road networks using graph con- volutional networks,

    J. Hu, C. Guo, B. Yang, and C. S. Jensen, “Stochastic weight completion for road networks using graph con- volutional networks,” in 2019 IEEE 35th international conference on data engineering (ICDE) . IEEE, 2019, pp. 1274–1285

  6. [5]

    An edge feature aware heterogeneous graph neural net- work model to support tax evasion detection,

    B. Shi, B. Dong, Y . Xu, J. Wang, Y . Wang, and Q. Zheng, “An edge feature aware heterogeneous graph neural net- work model to support tax evasion detection,” Expert Systems with Applications , vol. 213, p. 118903, 2023

  7. [6]

    Tax evasion detection with fbne-pu algorithm based on pncgcn and pu learning,

    Y . Gao, B. Shi, B. Dong, Y . Wang, L. Mi, and Q. Zheng, “Tax evasion detection with fbne-pu algorithm based on pncgcn and pu learning,” IEEE Transactions on Knowl- edge and Data Engineering , vol. 35, no. 1, pp. 931–944, 2021

  8. [7]

    Spectral networks and locally connected networks on graphs,

    J. Bruna, W. Zaremba, A. Szlam, and Y . LeCun, “Spectral networks and locally connected networks on graphs,” arXiv preprint arXiv:1312.6203 , 2013

Show all 48 references
  1. [8]

    Semi-supervised classifica- tion with graph convolutional networks,

    T. N. Kipf and M. Welling, “Semi-supervised classifica- tion with graph convolutional networks,” arXiv preprint arXiv:1609.02907, 2016

  2. [9]

    How powerful are graph neural networks?

    K. Xu, W. Hu, J. Leskovec, and S. Jegelka, “How powerful are graph neural networks?” arXiv preprint arXiv:1810.00826, 2018

  3. [10]

    Graph based semi- supervised learning with convolution neural networks to classify crisis related tweets,

    F. Alam, S. Joty, and M. Imran, “Graph based semi- supervised learning with convolution neural networks to classify crisis related tweets,” in Proceedings of the international AAAI conference on web and social media , vol. 12, 2018

  4. [11]

    Grale: Designing networks for graph learning,

    J. Halcrow, A. Mosoi, S. Ruth, and B. Perozzi, “Grale: Designing networks for graph learning,” in Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining , 2020, pp. 2523– 2532

  5. [12]

    Data augmentation for graph neural networks,

    T. Zhao, Y . Liu, L. Neves, O. Woodford, M. Jiang, and N. Shah, “Data augmentation for graph neural networks,” in Proceedings of the aaai conference on artificial intel- ligence, vol. 35, 2021, pp. 11 015–11 023

  6. [13]

    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 Proceedings of the AAAI conference on artificial intelligence, vol. 37, 2023, pp. 8914–8922

  7. [14]

    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 Proceedings of the AAAI Conference on Artificial Intelligence, vol. 36, 2022, pp. 7603–7611

  8. [15]

    Cldg: Contrastive learning on dynamic graphs,

    Y . Xu, B. Shi, T. Ma, B. Dong, H. Zhou, and Q. Zheng, “Cldg: Contrastive learning on dynamic graphs,” in 2023 IEEE 39th International Conference on Data Engineer- ing (ICDE). IEEE, 2023, pp. 696–707

  9. [16]

    Nrgnn: Learning a label noise resistant graph neural network on sparsely and noisily labeled graphs,

    E. Dai, C. Aggarwal, and S. Wang, “Nrgnn: Learning a label noise resistant graph neural network on sparsely and noisily labeled graphs,” in Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining, 2021, pp. 227–236

  10. [17]

    Robust triple-matrix-recovery-based auto- weighted label propagation for classification,

    H. Zhang, Z. Zhang, M. Zhao, Q. Ye, M. Zhang, and M. Wang, “Robust triple-matrix-recovery-based auto- weighted label propagation for classification,” IEEE Transactions on Neural Networks and Learning Systems , vol. 31, no. 11, pp. 4538–4552, 2020

  11. [18]

    Robust training of graph neural networks via noise governance,

    S. Qian, H. Ying, R. Hu, J. Zhou, J. Chen, D. Z. Chen, and J. Wu, “Robust training of graph neural networks via noise governance,” in Proceedings of the Sixteenth ACM International Conference on Web Search and Data Mining, 2023, pp. 607–615

  12. [19]

    Gnn cleaner: Label cleaner for graph structured data,

    J. Xia, H. Lin, Y . Xu, C. Tan, L. Wu, S. Li, and S. Z. Li, “Gnn cleaner: Label cleaner for graph structured data,” IEEE Transactions on Knowledge and Data Engineering, 2023

  13. [20]

    Learning on graphs under label noise,

    J. Yuan, X. Luo, Y . Qin, Y . Zhao, W. Ju, and M. Zhang, “Learning on graphs under label noise,” in IEEE Inter- national Conference on Acoustics, Speech and Signal Processing (ICASSP) 2023-2023. IEEE, 2023, pp. 1–5

  14. [21]

    Dividemix: Learning with noisy labels as semi-supervised learning,

    J. Li, R. Socher, and S. C. Hoi, “Dividemix: Learning with noisy labels as semi-supervised learning,” arXiv preprint arXiv:2002.07394, 2020

  15. [22]

    Deep self-learning from noisy labels,

    J. Han, P. Luo, and X. Wang, “Deep self-learning from noisy labels,” in Proceedings of the IEEE/CVF interna- tional conference on computer vision , 2019, pp. 5138– 5147

  16. [23]

    Joint optimization framework for learning with noisy labels,

    D. Tanaka, D. Ikami, T. Yamasaki, and K. Aizawa, “Joint optimization framework for learning with noisy labels,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2018, pp. 5552–5560

  17. [25]

    Partial label learning with batch label correction,

    Y . Yan and Y . Guo, “Partial label learning with batch label correction,” in proceedings of the AAAI conference on artificial intelligence, vol. 34, no. 04, 2020, pp. 6575– 6582

  18. [26]

    Making deep neural networks robust to label noise: A loss correction approach,

    G. Patrini, A. Rozza, A. Krishna Menon, R. Nock, and L. Qu, “Making deep neural networks robust to label noise: A loss correction approach,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 1944–1952

  19. [27]

    Co-teaching: Robust training of deep neural networks with extremely noisy labels,

    B. Han, Q. Yao, X. Yu, G. Niu, M. Xu, W. Hu, I. Tsang, and M. Sugiyama, “Co-teaching: Robust training of deep neural networks with extremely noisy labels,” Advances in neural information processing systems , vol. 31, 2018

  20. [28]

    How does disagreement help generaliza- tion against label corruption?

    X. Yu, B. Han, J. Yao, G. Niu, I. Tsang, and M. Sugiyama, “How does disagreement help generaliza- tion against label corruption?” in International Confer- ence on Machine Learning . PMLR, 2019, pp. 7164– 7173

  21. [29]

    Generalized cross entropy loss for training deep neural networks with noisy la- bels,

    Z. Zhang and M. Sabuncu, “Generalized cross entropy loss for training deep neural networks with noisy la- bels,” Advances in neural information processing sys- tems, vol. 31, 2018

  22. [30]

    Learning with instance-dependent label noise: A sample sieve approach,

    H. Cheng, Z. Zhu, X. Li, Y . Gong, X. Sun, and Y . Liu, “Learning with instance-dependent label noise: A sample sieve approach,” arXiv preprint arXiv:2010.02347, 2020

  23. [31]

    Class2simi: A noise reduction per- spective on learning with noisy labels,

    S. Wu, X. Xia, T. Liu, B. Han, M. Gong, N. Wang, H. Liu, and G. Niu, “Class2simi: A noise reduction per- spective on learning with noisy labels,” in International Conference on Machine Learning . PMLR, 2021, pp. 11 285–11 295

  24. [32]

    Instance-dependent label- noise learning with manifold-regularized transition ma- trix estimation,

    D. Cheng, T. Liu, Y . Ning, N. Wang, B. Han, G. Niu, X. Gao, and M. Sugiyama, “Instance-dependent label- noise learning with manifold-regularized transition ma- trix estimation,” in Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, 2022, pp. ...

  25. [33]

    Peer loss functions: Learning from noisy labels without knowing noise rates,

    Y . Liu and H. Guo, “Peer loss functions: Learning from noisy labels without knowing noise rates,” in Interna- tional conference on machine learning . PMLR, 2020, pp. 6226–6236

  26. [34]

    Robust loss functions under label noise for deep neural networks,

    A. Ghosh, H. Kumar, and P. S. Sastry, “Robust loss functions under label noise for deep neural networks,” in Proceedings of the AAAI conference on artificial intelligence, vol. 31, 2017

  27. [35]

    Learning from noisy labels with complementary loss functions,

    D.-B. Wang, Y . Wen, L. Pan, and M.-L. Zhang, “Learning from noisy labels with complementary loss functions,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 35, 2021, pp. 10 111–10 119

  28. [36]

    Neural message passing for quantum chem- istry,

    J. Gilmer, S. S. Schoenholz, P. F. Riley, O. Vinyals, and G. E. Dahl, “Neural message passing for quantum chem- istry,” in International conference on machine learning . PMLR, 2017, pp. 1263–1272

  29. [37]

    Dual t: Reducing estimation error for transition matrix in label-noise learning,

    Y . Yao, T. Liu, B. Han, M. Gong, J. Deng, G. Niu, and M. Sugiyama, “Dual t: Reducing estimation error for transition matrix in label-noise learning,” Advances in neural information processing systems , vol. 33, pp. 7260–7271, 2020

  30. [38]

    Part-dependent label noise: Towards instance-dependent label noise,

    X. Xia, T. Liu, B. Han, N. Wang, M. Gong, H. Liu, G. Niu, D. Tao, and M. Sugiyama, “Part-dependent label noise: Towards instance-dependent label noise,” Advances in Neural Information Processing Systems , vol. 33, pp. 7597–7610, 2020

  31. [39]

    Estimating instance-dependent bayes-label transition matrix using a deep neural network,

    S. Yang, E. Yang, B. Han, Y . Liu, M. Xu, G. Niu, and T. Liu, “Estimating instance-dependent bayes-label transition matrix using a deep neural network,” in In- ternational Conference on Machine Learning . PMLR, 2022, pp. 25 302–25 312

  32. [40]

    L dmi: A novel information-theoretic loss function for training deep nets robust to label noise,

    Y . Xu, P. Cao, Y . Kong, and Y . Wang, “L dmi: A novel information-theoretic loss function for training deep nets robust to label noise,” Advances in neural information processing systems, vol. 32, 2019

  33. [41]

    Classification with noisy labels by importance reweighting,

    T. Liu and D. Tao, “Classification with noisy labels by importance reweighting,” IEEE Transactions on pattern analysis and machine intelligence , vol. 38, no. 3, pp. 447–461, 2015

  34. [42]

    Birds of a feather: Homophily in social networks,

    M. McPherson, L. Smith-Lovin, and J. M. Cook, “Birds of a feather: Homophily in social networks,” Annual review of sociology, vol. 27, no. 1, pp. 415–444, 2001

  35. [43]

    Hierarchical grammar-induced geometry for data-efficient molecular property predic- tion,

    M. Guo, V . Thost, S. W. Song, A. Balachandran, P. Das, J. Chen, and W. Matusik, “Hierarchical grammar-induced geometry for data-efficient molecular property predic- tion,” in International Conference on Machine Learning . PMLR, 2023, pp. 12 055–12 076

  36. [44]

    Collective classification in network data,

    P. Sen, G. Namata, M. Bilgic, L. Getoor, B. Galligher, and T. Eliassi-Rad, “Collective classification in network data,” AI magazine, vol. 29, no. 3, pp. 93–93, 2008

  37. [45]

    Igb: Addressing the gaps in label- ing, features, heterogeneity, and size of public graph datasets for deep learning research,

    A. Khatua, V . S. Mailthody, B. Taleka, T. Ma, X. Song, and W.-m. Hwu, “Igb: Addressing the gaps in label- ing, features, heterogeneity, and size of public graph datasets for deep learning research,” arXiv preprint arXiv:2302.13522, 2023

  38. [46]

    Open graph benchmark: Datasets for machine learning on graphs,

    W. Hu, M. Fey, M. Zitnik, Y . Dong, H. Ren, B. Liu, M. Catasta, and J. Leskovec, “Open graph benchmark: Datasets for machine learning on graphs,” Advances in neural information processing systems , vol. 33, pp. 22 118–22 133, 2020

  39. [47]

    Learning from massive noisy labeled data for image classification,

    T. Xiao, T. Xia, Y . Yang, C. Huang, and X. Wang, “Learning from massive noisy labeled data for image classification,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2015, pp. 2691–2699

  40. [48]

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

    Q. Li, Z. Han, and X.-M. Wu, “Deeper insights into graph convolutional networks for semi-supervised learning,” in Proceedings of the AAAI conference on artificial intelligence, vol. 32, 2018

Pith tools

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