Pith. sign in

REVIEW 4 major objections 5 minor 51 references

Efficient Adaptive Label Refinement for Label Noise Learning

T0 review · 4 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that a two-term loss—cross-entropy against temporally ensembled soft labels plus a prediction-entropy regularizer—can avoid memorizing wrong labels while learning clean samples thoroughly, outperforming more complex…

desk verdict Simple, solid label-refurbishment variant whose empirical core holds up on CIFAR, but the theory is shaky and the SOTA claim overreaches. read the letter →

arxiv 2502.00386 v1 pith:C6WZA5W6 submitted 2025-02-01 cs.CV

Wenzhen Zhang , Debo Cheng , Guangquan Lu , Bo Zhou , Jiaye Li , Shichao Zhang This is my paper · ORCID
classification cs.CV
keywords labelnoiserefinementtemporalensemblingentropyregularizationsoftlabelsdeepneuralnetworksimageclassificationnoisy
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

Adaptive Label Refinement (ALR) is a training procedure that separates two problems usually solved together: not memorizing incorrect annotations, and learning as much as possible from correct ones. It replaces one-hot labels with soft labels formed by a temporal ensemble of the model's own past predictions, which dilutes the influence of any single noisy annotation. To stop those soft labels from making the model lazy on clean data, it adds a prediction-entropy term that gradually sharpens high-confidence soft labels back toward one-hot form. A gradient analysis in the paper shows that the combined loss pushes harder on confident samples while leaving low-confidence samples less forced, and the authors argue this is why the method can skip explicit clean/noisy partitioning. Experiments on CIFAR-10/100 with injected noise and on ANIMAL-10N, Clothing1M, and WebVision are reported as showing that this simple combination matches or beats the compared methods.

What carries the argument

The load-bearing object is the combined loss $\mathcal{L}_{\mathrm{ALR}} = \mathcal{L}_{\mathrm{ce}} + \lambda \mathcal{L}_e$, whose two terms are assigned two different jobs. The temporal-ensemble target $t^{(e)}$ defined by Eq. (3) is the label-refinement mechanism: it blends the original noisy one-hot label with a running average of the model's predictions, so any single wrong annotation is diluted over time. The entropy term $\mathcal{L}_e$ is the hardening mechanism: because its gradient contains $\log p[u] + H(p)$, the regularizer changes sign at a confidence threshold $\varepsilon$, so confident predictions are pushed toward one-hot while uncertain predictions are left alone. The theorem's sign analysis of this gradient is what connects the two terms: it shows that the entropy regularizer does not interfere with the soft-label dilution for low-confidence samples, and it supplies the extra learning pressure on confident samples that plain label refurbishment lacks.

What would settle it

A concrete check: on CIFAR-10 with 40% asymmetric noise (for example, the paper's cat-to-dog and bird-to-airplane flips), record each mislabeled sample's top-1 confidence $p[u]$ during the refinement phase. If a large fraction of mislabeled samples exceed the hardening threshold $\varepsilon$ and the model's accuracy is no better than the same model without the entropy term, then the claim that confident samples are clean is not doing the work claimed.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central claim is that the loss $\mathcal{L}_{\mathrm{ALR}} = \mathcal{L}_{\mathrm{ce}} + \lambda \mathcal{L}_e$—where the cross-entropy term is computed against temporally ensembled soft labels $t^{(e)} = \alpha t^{(e-1)} + (1-\alpha)p^{(e)}$ after a warm-up of $m$ epochs, and $\mathcal{L}_e = -\sum_i \sum_k p_i[k]\log p_i[k]$ is the entropy of the predictions—simultaneously resists memorizing wrong labels and learns clean labels thoroughly. The paper derives the per-sample gradient with respect to the top predicted logit $z[u]$ as $\partial \ell_{\mathrm{ALR}}/\partial z[u] = p[u]\bigl(1-\lambda(\log p[u]+H(p))\bigr)-t[u]$, and proves there is a unique confidence threshold $\varepsilon$ such that for $p[u]>\varepsilon$ the ALR gradient is smaller than the plain cross-entropy gradient while for $p[u]\le \varepsilon$ it is not. This is interpreted as giving high-confidence—hence, the paper assumes, clean—samples stronger learning pressure while not forcing low-confidence samples toward possibly wrong labels. The paper reports that this mechanism outperforms robust losses, regularizers, sample-selection methods, and label-refurbishment baselines on the five tested benchmarks.

Load-bearing premise

The method rests on the premise that high-confidence predictions are almost always correct, so hardening them can only help; when class imbalance or asymmetric noise produces confidently wrong predictions, that premise fails and the entropy term would harden incorrect soft labels.

Editorial extensions

If this is right

  • In the paper's experiments, ALR outperforms the compared methods under both symmetric and asymmetric noise, including asymmetric settings where small-loss selection criteria are known to struggle.
  • A classifier can adopt ALR by changing only the loss and the label-update rule; no noise-rate estimate, clean auxiliary set, or explicit sample selection is required.
  • On clean data, the soft-label target and entropy term act as a regularizer, and the paper's ablations report that ALR still slightly beats plain cross-entropy on CIFAR-10 and CIFAR-100 without injected noise.
  • The hyperparameters $m$, $\alpha$, and $\lambda$ are reported to be stable across datasets, with the main caution being that a very large momentum $\alpha$ hurts CIFAR-100 under asymmetric noise.

Reading between the lines

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

  • Beyond the paper, the gradient formula suggests a family of confidence-gated losses: any regularizer whose derivative changes sign at a confidence threshold should reproduce the harden-confident, leave-uncertain behaviour, so the entropy term may be replaceable by cheaper or better-calibrated alternatives.
  • A stress test the paper does not run is class-imbalanced noise combined with asymmetric flips; the authors' own explanation for the WebVision shortfall identifies confidence on rare classes as the weak point, so a dataset with severe imbalance and high noise would show whether the high-confidence-equals-clean premise can be sustained.
  • One could also apply the same temporal-ensembled soft labels plus entropy hardening to semi-supervised or partial-label settings, where pseudo-labels are refined iteratively; this is an extension the paper does not claim.
  • A practical follow-up would be to log the threshold $\varepsilon$ and the fraction of mislabeled samples whose confidence crosses it during training; if that fraction grows, a confidence-calibration or abstention mechanism would be needed.
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 / 5 minor

Summary. The paper proposes Adaptive Label Refinement (ALR), a training method for learning with noisy labels. After a warm-up phase, ALR replaces the original hard labels with temporally ensembled soft labels (Eq. 3) and adds a negative entropy regularizer (Eq. 5), training with L_ALR = L_CE + lambda * L_e. The authors argue that soft-label refurbishment reduces memorization of noisy labels while the entropy term progressively 'hardens' high-confidence targets, thereby improving learning from clean samples without explicit sample selection. The method is evaluated on CIFAR-10 and CIFAR-100 under symmetric and asymmetric noise and on ANIMAL-10N, Clothing1M, and mini-WebVision, with ablations and hyperparameter sensitivity experiments. The central claim is that ALR outperforms state-of-the-art methods while being simpler and requiring no noise-rate information or auxiliary clean data.

Significance. If the empirical results are taken at face value, the paper offers a simple and generally applicable recipe for label-noise robustness, with a plausible decoupling of noise avoidance and clean-sample learning. Strengths include fixed hyperparameters across the CIFAR noise conditions, ablations showing that both the label-refinement and entropy components contribute, and results on several real-world noisy datasets. The comparison is against external baselines, and no reported test number is fit by the method, so the circularity burden is low. However, the 'outperforms state-of-the-art' claim is weakened by the WebVision results, where ELR scores higher on the primary metric, and by the absence of DivideMix, a standard strong baseline. The theoretical analysis in Section 4.3 is also not as supportive as claimed: it establishes a gradient comparison for high-confidence predictions, not for clean samples. These issues are fixable but currently affect the paper's headline claims.

major comments (4)
  1. [§4.3, Eq. (7)-(8)] Theorem 1 as used in the text does not establish that ALR prioritizes clean samples. The inequality ∂ℓ_ALR/∂z[u] < p[u] − t[u] holds for every high-confidence prediction p[u] > ε, regardless of whether the predicted class is the true class, the noisy label, or neither. The entropy term depends only on p, so it cannot distinguish clean labels from noisy ones. Under class imbalance (acknowledged for WebVision) or asymmetric noise, a confidently wrong prediction satisfies the same condition, and the gradient then hardens the incorrect belief. Because t(e) in Eq. (3) tracks p(e) for consistently confident samples, the cross-entropy term does not reliably counteract this effect. No experiment in the paper measures the correlation between prediction confidence and label correctness, so the premise 'high-confidence samples are typically clean' is unverified. The theorem's conclusion should be restated as a property of high-confidence predictions, and the method's behavior under confident-but-wrong predictions should be examined empirically.
  2. [§4.3, Eq. (17)-(18)] The proof of Theorem 1 contains formal errors. Eq. (18) writes f'(p[u]) = 1/p[u] − (log p[k]+1) with the index k in place of u, and the derivative ∂H(p)/∂p[u] is computed as if the other probabilities were fixed even though the probability vector satisfies Σ p[k] = 1. As written, the existence and uniqueness of the threshold ε is not rigorously established; at best, the argument is a pointwise statement for a fixed distribution. The proof should be corrected or the theorem weakened accordingly before the result is cited as a guarantee. The final gradient formula in Eq. (7) is nevertheless correct, which suggests the issue is in the proof presentation rather than in the underlying calculation.
  3. [Abstract; §5.3, Table 7] The abstract's claim that 'ALR outperforms state-of-the-art methods' is contradicted by the authors' own Table 7. On WebVision top-1, ELR reaches 76.26 while ALR reaches 74.72; on top-5, ELR reaches 91.26 versus ALR's 90.84. The text acknowledges the result as 'slightly inferior' and attributes it to class imbalance, but no experiment in the paper tests that attribution. The abstract and conclusion should be qualified to reflect that ALR is competitive with, but not uniformly superior to, the compared methods on this benchmark.
  4. [§2, §5.2-5.3, Tables 2-7] DivideMix, which the authors cite and describe in Section 2, is a standard strong baseline for the CIFAR and real-world label-noise benchmarks used in the paper, yet it is absent from all comparison tables. Without this baseline, the statement that ALR outperforms state-of-the-art methods is not supported by the evidence presented. The authors should either add DivideMix to the comparison (using published numbers under the same protocol) or explicitly restrict their superiority claims to the methods actually compared.
minor comments (5)
  1. [§4.2, Eq. (3) and Algorithm 1] The warm-up condition is inconsistent: Eq. (3) states that the warm-up phase is e ≤ m and refinement starts at e > m, while Algorithm 1 (line 3) switches to refinement when e < m. Please align the equation and the pseudocode.
  2. [§4.2, Eq. (4)] The sentence 'Substituting Eq. (3) into Eq. (3) gives the loss function' should read 'Substituting Eq. (3) into Eq. (1) gives the loss function.'
  3. [§5.6 and Figures 7-9] The sensitivity analysis swaps the roles of α and λ. Section 5.6.1 tests the entropy-loss weight but labels it α and concludes 'we set α to 0.2'; Section 5.6.3 tests the temporal momentum but labels it λ and concludes 'we fix λ at 0.9.' Elsewhere in the paper α is the temporal momentum and λ is the entropy-loss weight. The same swap appears in the captions of Figures 7 and 9. In addition, Section 5.6.2 and Figure 8 call m the 'momentum hyperparameter' although m denotes the number of warm-up epochs.
  4. [§3.1, Eq. (1)] The notation ˜yi ∈ (0,1)^K is inaccurate for a one-hot vector; the entries are 0 or 1, so the correct set is {0,1}^K (or the interval notation should be clarified).
  5. [§4.3, proof of Theorem 1] The proof contains typos: Eq. (11) differentiates t[k] instead of p[k], Eq. (13) contains a duplicated factor, and the line 'By combining Eq. (3) and (4)' refers to equations in the proof rather than to the main-text Eq. (3) and Eq. (4). These should be corrected for readability.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; ALR is an empirical training recipe with fixed hyperparameters and external benchmarks, and Theorem 1 is a conditional gradient computation rather than a self-referential fit.

full rationale

ALR is a proposed training recipe rather than a derivation whose predictions are re-fit from the reported test numbers. The objective L_ALR = L_ce + lambda*L_e (Eq. 6) is defined independently of the test accuracies; the hyperparameters (alpha=0.9, lambda=0.2, warm-up m=30) are fixed across CIFAR noise settings, sensitivity analyses are reported, and the comparisons use baseline numbers from the respective original papers. Theorem 1 is an exact gradient computation: Eq. (7) follows from calculus, and the inequality in Eq. (8) is a mathematical consequence, not a quantity fitted to the benchmark results. The subsequent statement that 'High-confidence samples are typically clean' is an unverified empirical premise and a correctness risk, especially under class imbalance, but it is not a circular definition: the method does not define clean samples as p[u]>epsilon, and the paper explicitly acknowledges the WebVision case where the premise weakens. The few self-citations (e.g., Liang et al. 2024 for WebVision imbalance) support background dataset facts and are not load-bearing in the derivation chain. No prediction or first-principles result in the paper is equivalent by construction to its inputs.

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

ALR's central claim rests on three hand-set hyperparameters (alpha, lambda, m) and on three domain assumptions about early learning, confidence-correctness correlation, and temporal ensemble reliability. No new physical entities or external fitted constants are introduced.

free parameters (3)
  • alpha (temporal momentum) = 0.9
    Controls soft-label update t(e)=alpha*t(e-1)+(1-alpha)*p(e) in Eq. (3); fixed at 0.9 based on sensitivity analysis in Section 5.6.3 (labeled there as lambda).
  • lambda (entropy loss weight) = 0.2
    Weight on entropy term in Eq. (6); fixed at 0.2 based on sensitivity in Section 5.6.1 (labeled there as alpha).
  • m (warm-up epochs) = 30
    Duration of CE training with original noisy labels before refinement; fixed at 30 based on Section 5.6.2.
assumptions (4)
  • domain assumption Early learning: DNNs fit clean samples before noisy samples.
    Invoked in Sections 3.1 and 4.1 to justify the warm-up phase; imported from Zhang et al. 2017 and Arpit et al. 2017.
  • domain assumption High-confidence predictions are typically correct.
    Stated in Section 4.3 to argue the entropy term hardens clean samples; not proven and can fail under class imbalance or asymmetric noise.
  • domain assumption Temporal ensembling targets t(e) are reliable estimates of true labels.
    Eq. (3) assumes the running average of historical predictions is a better target than the original noisy label; depends on the quality of the warm-up model.
  • standard math Softmax gradient identities and the intermediate value theorem.
    Used in the Theorem 1 proof; the boundary and monotonicity steps are applied incorrectly or without full justification.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Efficient Adaptive Label Refinement for Label Noise Learning." pith.science (2026). https://pith.science/paper/C6WZA5W6

@misc{pith2026250200386,
  author       = {Pith},
  title        = {Pith review of: Efficient Adaptive Label Refinement for Label Noise Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/C6WZA5W6}},
  note         = {Machine review of arXiv:2502.00386}
}
read the original abstract

Deep neural networks are highly susceptible to overfitting noisy labels, which leads to degraded performance. Existing methods address this issue by employing manually defined criteria, aiming to achieve optimal partitioning in each iteration to avoid fitting noisy labels while thoroughly learning clean samples. However, this often results in overly complex and difficult-to-train models. To address this issue, we decouple the tasks of avoiding fitting incorrect labels and thoroughly learning clean samples and propose a simple yet highly applicable method called Adaptive Label Refinement (ALR). First, inspired by label refurbishment techniques, we update the original hard labels to soft labels using the model's predictions to reduce the risk of fitting incorrect labels. Then, by introducing the entropy loss, we gradually `harden' the high-confidence soft labels, guiding the model to better learn from clean samples. This approach is simple and efficient, requiring no prior knowledge of noise or auxiliary datasets, making it more accessible compared to existing methods. We validate ALR's effectiveness through experiments on benchmark datasets with artificial label noise (CIFAR-10/100) and real-world datasets with inherent noise (ANIMAL-10N, Clothing1M, WebVision). The results show that ALR outperforms state-of-the-art methods.

Figures

Figures reproduced from arXiv: 2502.00386 by the authors.

Figure 1
Figure 1. t-SNE visualizations in plots (a) and (b) illustrate the relationships among [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Plots (a) and (b) illustrate the training and testing accuracies of DNNs employing [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. The architecture of the ALR training framework. In the warm-up phase, the [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: (a) and (b) illustrate the model’s memorization behavior on clean and mislabeled samples using cross-entropy loss (CE). In contrast, Fig￾ure 4 (c) and (d) show the model’s memorization behavior on clean and mislabeled samples using ALR. It is clear that ALR significant…
Figure 4
Figure 4. Figure 4: We studied the memorization phenomenon of DNNs on CIFAR-10 using CE and [PITH_FULL_IMAGE:figures/full_fig_p025_4.png]
Figure 5
Figure 5. Figure 5: 2D t-SNE visualization of clean samples in the CIFAR-10 training set for the [PITH_FULL_IMAGE:figures/full_fig_p026_5.png]
Figure 6
Figure 6. Figure 6: Models were trained using ALR on the CIFAR-10 training set with different [PITH_FULL_IMAGE:figures/full_fig_p027_6.png]
Figure 7
Figure 7. Figure 7: Impact of the entropy loss regularization weight coefficient [PITH_FULL_IMAGE:figures/full_fig_p029_7.png]
Figure 8
Figure 8. Figure 8: Impact of the momentum hyperparameter m on test classification accuracy on CIFAR-10 (C10) and CIFAR-100 (C100) under 40% symmetric and asymmetric label noise. 5.6.3. Sensitivity Analysis of the momentum hyperparameter λ To evaluate the influence of the momentum paramet…
Figure 9
Figure 9. Figure 9: Impact of the warm-up phase of λ epochs on test classification accuracy on CIFAR-10 (C10) and CIFAR-100 (C100) under 40% symmetric and asymmetric label noise. 6. Conclusion We propose Adaptive Label Refinement (ALR) to address the challenge of label noise in classifica…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 47 canonical work pages

  1. [1]

    , author Ortego, D

    author Arazo, E. , author Ortego, D. , author Albert, P. , author O’Connor, N. , author McGuinness, K. , year 2019 . title Unsupervised label noise modeling and loss correction , in: booktitle International conference on machine learning , organization PMLR . pp. pages 312--321

    work page 2019

  2. [2]

    , author Jastrz e bski, S

    author Arpit, D. , author Jastrz e bski, S. , author Ballas, N. , author Krueger, D. , author Bengio, E. , author Kanwal, M.S. , author Maharaj, T. , author Fischer, A. , author Courville, A. , author Bengio, Y. , et al., year 2017 . title A closer look at memorization in deep networks , in: booktitle International conference on machine learning , organiz...

    work page 2017

  3. [3]

    , author Liao, B.B

    author Chen, P. , author Liao, B.B. , author Chen, G. , author Zhang, S. , year 2019 . title Understanding and utilizing deep neural networks trained with noisy labels , in: booktitle International conference on machine learning , organization PMLR . pp. pages 1062--1070

    work page 2019

  4. [4]

    , author Shen, X

    author Chen, Y. , author Shen, X. , author Hu, S.X. , author Suykens, J.A. , year 2021 . title Boosting co-teaching with compression regularization for label noise , in: booktitle Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp. pages 2688--2692

    work page 2021

  5. [5]

    , author Ning, Y

    author Cheng, D. , author Ning, Y. , author Wang, N. , author Gao, X. , author Yang, H. , author Du, Y. , author Han, B. , author Liu, T. , year 2022 . title Class-dependent label-noise learning with cycle-consistency regularization . journal Advances in Neural Information Processing Systems volume 35 , pages 11104--11116

    work page 2022

  6. [6]

    , author Kumar, H

    author Ghosh, A. , author Kumar, H. , author Sastry, P.S. , year 2017 . title Robust loss functions under label noise for deep neural networks , in: booktitle Proceedings of the AAAI conference on artificial intelligence , pp. pages 1919--1925

    work page 2017

  7. [7]

    , author Ding, Y

    author Gong, C. , author Ding, Y. , author Han, B. , author Niu, G. , author Yang, J. , author You, J. , author Tao, D. , author Sugiyama, M. , year 2022 . title Class-wise denoising for robust learning under label noise . journal IEEE Transactions on Pattern Analysis and Machine Intelligence volume 45 , pages 2835--2848

    work page 2022

  8. [8]

    , author Bisht, N

    author Gong, X. , author Bisht, N. , author Xu, G. , year 2024 . title Does label smoothing help deep partial label learning? , in: booktitle Forty-first International Conference on Machine Learning

    work page 2024

Show all 51 references
  1. [9]

    , author Wang, W

    author Gui, X.J. , author Wang, W. , author Tian, Z.H. , year 2021 . title Towards understanding deep learning from noisy labels with small-loss criterion , in: booktitle Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence , pp. pages 2469--2475

    2021

  2. [10]

    , author Yao, Q

    author Han, B. , author Yao, Q. , author Yu, X. , author Niu, G. , author Xu, M. , author Hu, W. , author Tsang, I. , author Sugiyama, M. , year 2018 . title Co-teaching: Robust training of deep neural networks with extremely noisy labels . journal Advances in neural informati...

    2018

  3. [11]

    , author Zhang, X

    author He, K. , author Zhang, X. , author Ren, S. , author Sun, J. , year 2016 . title Deep residual learning for image recognition , in: booktitle Proceedings of the IEEE conference on computer vision and pattern recognition , pp. pages 770--778

    2016

  4. [12]

    , author Zhou, Z

    author Jiang, L. , author Zhou, Z. , author Leung, T. , author Li, L.J. , author Fei-Fei, L. , year 2018 . title Mentornet: Learning data-driven curriculum for very deep neural networks on corrupted labels , in: booktitle International conference on machine learning , organiza...

    2018

  5. [13]

    , author Yan, Y

    author Jiang, R. , author Yan, Y. , author Xue, J.H. , author Wang, B. , author Wang, H. , year 2022 . title When sparse neural network meets label noise learning: A multistage learning framework . journal IEEE Transactions on Neural Networks and Learning Systems volume 35 , p...

    2022

  6. [14]

    , author Yim, J

    author Kim, Y. , author Yim, J. , author Yun, J. , author Kim, J. , year 2019 . title Nlnl: Negative learning for noisy labels , in: booktitle Proceedings of the IEEE/CVF international conference on computer vision , pp. pages 101--110

    2019

  7. [15]

    , author Hinton, G

    author Krizhevsky, A. , author Hinton, G. , et al., year 2009 . title Learning multiple layers of features from tiny images

    2009

  8. [16]

    , author Hertz, J

    author Krogh, A. , author Hertz, J. , year 1991 . title A simple weight decay can improve generalization . journal Advances in neural information processing systems volume 4 , pages 950--957

    1991

  9. [17]

    , author Li, X

    author Li, D. , author Li, X. , author Gan, Z. , author Li, Q. , author Qu, B. , author Wang, J. , year 2024 . title Rethinking the impact of noisy labels in graph classification: A utility and privacy perspective . journal Neural Networks , pages 106919

    2024

  10. [18]

    , author Socher, R

    author Li, J. , author Socher, R. , author Hoi, S.C. , year 2020 . title Dividemix: Learning with noisy labels as semi-supervised learning , in: booktitle International Conference on Learning Representations

    2020

  11. [19]

    , author Wang, L

    author Li, W. , author Wang, L. , author Li, W. , author Agustsson, E. , author Van Gool, L. , year 2017 . title Webvision database: Visual learning and understanding from web data . journal arXiv preprint arXiv:1708.02862

    2017 arXiv

  12. [20]

    , author Xia, X

    author Li, X.C. , author Xia, X. , author Zhu, F. , author Liu, T. , author Zhang, X.Y. , author Liu, C.L. , year 2023 . title Dynamics-aware loss for learning with label noise . journal Pattern Recognition volume 144 , pages 109835

    2023

  13. [21]

    , author Zhang, S

    author Liang, R. , author Zhang, S. , author Zhang, W. , author Zhang, G. , author Tang, J. , year 2024 . title Nonlocal hybrid network for long-tailed image classification . journal ACM Transactions on Multimedia Computing, Communications and Applications volume 20 , pages 1--22

    2024

  14. [22]

    , author Niles-Weed, J

    author Liu, S. , author Niles-Weed, J. , author Razavian, N. , author Fernandez-Granda, C. , year 2020 . title Early-learning regularization prevents memorization of noisy labels . journal Advances in neural information processing systems volume 33 , pages 20331--20342

    2020

  15. [23]

    , author Zhu, Z

    author Liu, S. , author Zhu, Z. , author Qu, Q. , author You, C. , year 2022 . title Robust training under label noise by over-parameterization , in: booktitle International Conference on Machine Learning , organization PMLR . pp. pages 14153--14172

    2022

  16. [24]

    , author Jiang, Y.G

    author Liu, W. , author Jiang, Y.G. , author Luo, J. , author Chang, S.F. , year 2011 . title Noise resistant graph ranking for improved web image search , in: booktitle CVPR 2011 , organization IEEE . pp. pages 849--856

    2011

  17. [25]

    , author He, W

    author Lu, Y. , author He, W. , year 2022 . title Selc: Self-ensemble label correction improves learning with noisy labels . journal IJCAI , pages 3278--3284

    2022

  18. [26]

    , author Xu, Z

    author Lu, Y. , author Xu, Z. , author He, W. , year 2023 . title Rethinking label refurbishment: model robustness under label noise , in: booktitle Proceedings of the AAAI Conference on Artificial Intelligence , pp. pages 15000--15008

    2023

  19. [27]

    , author Mummadi, C

    author Nguyen, T. , author Mummadi, C. , author Ngo, T. , author Beggel, L. , author Brox, T. , year 2020 . title Self: learning to filter noisy labels with self-ensembling , in: booktitle International Conference on Learning Representations (ICLR)

    2020

  20. [28]

    , author Rozza, A

    author Patrini, G. , author Rozza, A. , author Krishna Menon, A. , author Nock, R. , author Qu, L. , year 2017 . title Making deep neural networks robust to label noise: A loss correction approach , in: booktitle Proceedings of the IEEE conference on computer vision and patter...

    2017

  21. [29]

    , author Zhang, T

    author Pleiss, G. , author Zhang, T. , author Elenberg, E. , author Weinberger, K.Q. , year 2020 . title Identifying mislabeled data using the area under the margin ranking . journal Advances in Neural Information Processing Systems volume 33 , pages 17044--17056

    2020

  22. [30]

    , author Lee, H

    author Reed, S.E. , author Lee, H. , author Anguelov, D. , author Szegedy, C. , author Erhan, D. , author Rabinovich, A. , year 2015 . title Training deep neural networks on noisy labels with bootstrapping , in: booktitle International Conference on Learning Representations

    2015

  23. [31]

    , author Guo, Z

    author Shi, X. , author Guo, Z. , author Li, K. , author Liang, Y. , author Zhu, X. , year 2023 . title Self-paced resistance learning against overfitting on noisy labels . journal Pattern Recognition volume 134 , pages 109080

    2023

  24. [32]

    , author Khoshgoftaar, T.M

    author Shorten, C. , author Khoshgoftaar, T.M. , year 2019 . title A survey on image data augmentation for deep learning . journal Journal of big data volume 6 , pages 1--48

    2019

  25. [33]

    , author Xie, Q

    author Shu, J. , author Xie, Q. , author Yi, L. , author Zhao, Q. , author Zhou, S. , author Xu, Z. , author Meng, D. , year 2019 . title Meta-weight-net: Learning an explicit mapping for sample weighting . journal Advances in neural information processing systems volume 32 , ...

    2019

  26. [34]

    , author Kim, M

    author Song, H. , author Kim, M. , author Lee, J.G. , year 2019 . title Selfie: Refurbishing unclean samples for robust deep learning , in: booktitle International conference on machine learning , organization PMLR . pp. pages 5907--5915

    2019

  27. [35]

    , author Hinton, G

    author Srivastava, N. , author Hinton, G. , author Krizhevsky, A. , author Sutskever, I. , author Salakhutdinov, R. , year 2014 . title Dropout: a simple way to prevent neural networks from overfitting . journal The journal of machine learning research volume 15 , pages 1929--1958

    2014

  28. [36]

    , author Vanhoucke, V

    author Szegedy, C. , author Vanhoucke, V. , author Ioffe, S. , author Shlens, J. , author Wojna, Z. , year 2016 . title Rethinking the inception architecture for computer vision , in: booktitle Proceedings of the IEEE conference on computer vision and pattern recognition , pp....

    2016

  29. [37]

    , author Ma, X

    author Wang, Y. , author Ma, X. , author Chen, Z. , author Luo, Y. , author Yi, J. , author Bailey, J. , year 2019 . title Symmetric cross entropy for robust learning with noisy labels , in: booktitle Proceedings of the IEEE/CVF international conference on computer vision , pp...

    2019

  30. [38]

    , author Han, B

    author Xia, X. , author Han, B. , author Zhan, Y. , author Yu, J. , author Gong, M. , author Gong, C. , author Liu, T. , year 2023 . title Combating noisy labels with sample selection by mining high-discrepancy examples , in: booktitle Proceedings of the IEEE/CVF international...

    2023

  31. [39]

    , author Liu, T

    author Xia, X. , author Liu, T. , author Han, B. , author Gong, C. , author Wang, N. , author Ge, Z. , author Chang, Y. , year 2021 . title Robust early-learning: Hindering the memorization of noisy labels , in: booktitle International conference on learning representations

    2021

  32. [40]

    , author Xia, T

    author Xiao, T. , author Xia, T. , author Yang, Y. , author Huang, C. , author Wang, X. , year 2015 . title Learning from massive noisy labeled data for image classification , in: booktitle Proceedings of the IEEE conference on computer vision and pattern recognition , pp. pag...

    2015

  33. [41]

    , author Gong, M

    author Yao, Y. , author Gong, M. , author Du, Y. , author Yu, J. , author Han, B. , author Zhang, K. , author Liu, T. , year 2023 . title Which is better for learning with noisy labels: the semi-supervised method or modeling label noise? , in: booktitle International conferenc...

    2023

  34. [42]

    , author Wu, J

    author Yi, K. , author Wu, J. , year 2019 . title Probabilistic end-to-end noise correction for learning with noisy labels , in: booktitle Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pp. pages 7017--7025

    2019

  35. [43]

    , author Liu, S

    author Yi, L. , author Liu, S. , author She, Q. , author McLeod, A.I. , author Wang, B. , year 2022 . title On learning contrastive representations for learning with noisy labels , in: booktitle Proceedings of the IEEE/CVF conference on computer vision and pattern recognition ...

    2022

  36. [44]

    , author Han, B

    author Yu, X. , author Han, B. , author Yao, J. , author Niu, G. , author Tsang, I. , author Sugiyama, M. , year 2019 . title How does disagreement help generalization against label corruption? , in: booktitle International conference on machine learning , organization PMLR . ...

    2019

  37. [45]

    , author Liu, T

    author Yu, X. , author Liu, T. , author Gong, M. , author Tao, D. , year 2018 . title Learning with biased complementary labels , in: booktitle Proceedings of the European conference on computer vision (ECCV) , pp. pages 68--83

    2018

  38. [46]

    , author Bengio, S

    author Zhang, C. , author Bengio, S. , author Hardt, M. , author Recht, B. , author Vinyals, O. , year 2017 . title Understanding deep learning requires rethinking generalization , in: booktitle International Conference on Learning Representations

    2017

  39. [47]

    , author Zhu, C

    author Zhang, S. , author Zhu, C. , author Li, H. , author Cai, J. , author Yang, L. , year 2024 . title Gradient-aware learning for joint biases: Label noise and class imbalance . journal Neural Networks volume 171 , pages 374--382

    2024

  40. [48]

    , author Zheng, S

    author Zhang, Y. , author Zheng, S. , author Wu, P. , author Goswami, M. , author Chen, C. , year 2021 . title Learning with feature-dependent label noise: A progressive approach , in: booktitle ICLR

    2021

  41. [49]

    , author Sabuncu, M

    author Zhang, Z. , author Sabuncu, M. , year 2018 . title Generalized cross entropy loss for training deep neural networks with noisy labels . journal Advances in neural information processing systems volume 31 , pages 8792--8802

    2018

  42. [50]

    , author Liu, X

    author Zhou, X. , author Liu, X. , author Wang, C. , author Zhai, D. , author Jiang, J. , author Ji, X. , year 2021 . title Learning with noisy labels via sparse regularization , in: booktitle Proceedings of the IEEE/CVF international conference on computer vision , pp. pages 72--81

    2021

  43. [51]

    , author Feng, L

    author Zhu, Y. , author Feng, L. , author Deng, Z. , author Chen, Y. , author Amor, R. , author Witbrock, M. , year 2024 . title Robust node classification on graph data with graph and label noise , in: booktitle Proceedings of the AAAI Conference on Artificial Intelligence , ...

    2024

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.