Pith. sign in

REVIEW 4 major objections 7 minor 55 references

COLUR: Confidence-Oriented Learning, Unlearning and Relearning with Noisy-Label Data for Model Restoration and Refinement

T0 review · 4 major / 7 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read A model degraded by noisy labels can be restored to better-than-original accuracy by unlearning high-confidence teacher-student disagreements and relearning from confidence-refined soft labels, with no need to know which labels are wrong.

desk verdict Real problem, plausible mechanism, and large high-noise margins — but the baseline comparison protocol is under-specified and some baseline numbers look like artifacts, so the headline claim needs verification. read the letter →

arxiv 2506.19496 v1 pith:GCFBDQKV submitted 2025-06-24 cs.LG

classification cs.LG MSC 68T0768T05
keywords modelrestorationandrefinementlearningwithnoisylabelsmachineunlearningteacher-studentco-traininglabelsmoothingmixupconfidencenoise
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

The paper claims that a trained model whose accuracy has collapsed after incremental training on noisy labels can be restored, and even improved beyond its original level, by a pipeline that unlearns the influence of suspicious labels and then relearns from soft, confidence-refined labels. It introduces COLUR, a teacher-student framework that never needs to be told which labels are wrong. On CIFAR-10 with 90% label noise, COLUR lifts accuracy from 30.04% to 80.34%; at 50% noise it reaches 87.30%, above the original model's 84.85%. Across four datasets and noise levels from 10% to 90%, COLUR reports the best restored accuracy among both learning-with-noisy-labels and machine-unlearning baselines. If correct, this makes retraining from scratch unnecessary for noise-damaged models.

What carries the argument

The load-bearing object is the joint confidence score $c(x;\theta_t,\theta_u)=\sqrt{c(x;\theta_t)\cdot c(x;\theta_u)}$, the geometric mean of teacher and student max-class probabilities on a sample where the two models disagree. Samples scoring above a threshold $\tau$ form the high-confidence noisy set that is unlearned by gradient ascent on a label-smoothed target. The same teacher-student agreement geometry, with the student replaced by the unlearned model, produces the high-confidence agreement set that is relearned with label smoothing, while low-confidence data are mixup-augmented with high-confidence data to create soft-label training pairs. This confidence geometry is what turns unlearning from a data-removal operation into a self-supervised label-refinement loop.

What would settle it

Take a model trained on clean data, then present it with an incremental set of clean but hard examples (for instance naturally ambiguous images or a mild distribution shift) where the teacher is confidently wrong on many labels; if COLUR's unlearning stage removes those high-confidence disagreements and final accuracy falls below a control that skips unlearning, the sample-selection premise is refuted.

Watch

Extended reading notes

Core claim

The paper's central claim is that model performance lost to noisy-label training can be actively undone rather than merely tolerated: a copy of the original model acts as teacher, the degraded model as student, and samples where both predict different labels with high confidence are treated as mislabeled and unlearned through label-smoothing gradient ascent. The unlearned model is then relearned on high-confidence agreed labels, smoothed to avoid overconfidence, and on mixup-augmented data that blends low-confidence and high-confidence soft predictions. The two steps alternate. In the paper's experiments this restores CIFAR-10 from 30.04% to 80.34% accuracy at 90% noise and from 65.06% to 87.30% at 50% noise, surpassing the original undegraded model's 84.85%; similar gains are reported on CIFAR-100, Flower-102, and Oxford-IIIT Pet. The claim is that this repair path works without the user identifying any noisy label in advance.

Load-bearing premise

The whole repair rests on the assumption that when the original teacher model and the degraded student both predict different labels with high confidence, the sample is mislabeled; if the original model is confidently wrong about a clean-but-hard example, the unlearning step removes correct knowledge and relearning can bake in the damage.

Editorial extensions

If this is right

  • COLUR is model-agnostic: it only needs a copy of the original model and the degraded model, so it can repair different backbone architectures without architectural changes.
  • The unlearning stage removes the need for a pre-specified forget set: noisy samples are discovered by teacher-student disagreement rather than given by the user.
  • At high noise levels, LNL baselines collapse on several datasets while COLUR retains a large fraction of accuracy, suggesting the unlearn-then-relearn loop is most valuable exactly where standard robust training fails.
  • Because COLUR surpasses the original model at moderate noise, the relearning stage is not just restoration but refinement: the model can end up better than it was before degradation.
  • MU baselines that are handed the true noisy-label set still underperform COLUR, implying that targeted unlearning alone is insufficient and the relearning step is essential.

Reading between the lines

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

  • The teacher-student disagreement criterion could be adapted to detect other forms of training-data corruption, such as poisoned examples, since it only requires confident disagreement rather than knowledge of the corruption type.
  • A testable extension is to apply COLUR under distribution shift between the original data and the incremental noisy data; the fixed teacher's authority would then decay, so replacing it with an iteratively updated teacher might extend the framework to non-stationary domains.
  • The ablation shows the unlearning module alone gives modest gains while the full pipeline gives the large improvement, suggesting most of the benefit comes from relearning on refined soft labels; a natural experiment is to test whether simple downweighting of disagreement samples replaces the unlearning step.
  • Reporting accuracy on standard test splits leaves open how the restored model behaves under harder or shifted test sets; an independent check on corrupted or out-of-distribution variants would clarify whether the model generalizes or re-fits the clean structure of the incremental data.
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 / 7 minor

Summary. The paper introduces COLUR, a framework for restoring a model whose performance has degraded after incremental training on a dataset with noisy labels. The pipeline has three stages. First, learning: the original model f(θ0) is trained or fine-tuned on D_u to produce a degraded model f(θu). Second, unlearning: teacher/student disagreement sets are extracted, where the teacher is a copy of f(θ0), and high-confidence disagreements are unlearned via label-smoothed gradient ascent. Third, relearning: high-confidence agreements are relearned with label smoothing, and low-confidence samples are mixed with high-confidence samples via Mixup using soft labels. The unlearning and relearning steps are iterated. Experiments on CIFAR-10, CIFAR-100, Flower-102 and Oxford-IIIT Pet under symmetric and asymmetric noise at noise ratios from 10% to 90% compare COLUR with nine LNL and four MU baselines. The headline result is that COLUR achieves the highest accuracy in every reported setting, e.g., restoring CIFAR-10 from 30.04% (degraded) to 80.34% at 90% noise and exceeding the original model's accuracy at 50% noise.

Significance. The LUR-inspired framing is a genuinely useful reframing: instead of training an LNL model from scratch, the paper aims to repair a deployed model by unlearning the influence of noisy labels and relearning with refined soft labels. If the reported results are reproducible under a fair protocol, this would be a practical contribution, because full retraining is often too expensive and MU methods alone cannot correct labels. The paper also provides an anonymized code link and hyper-parameter tables, which aids reproducibility. The main weakness is that the empirical evidence, as currently presented, does not yet support the 'consistently outperforms SOTA' claim: the baseline-adaptation protocol is unspecified, several baseline numbers are difficult to reconcile with known LNL behavior, and no variance information is provided. The core idea is defensible and the defects are addressable, so the appropriate response is a major revision rather than rejection.

major comments (4)
  1. [Section 4.1, Tables 1 and 6] The protocol by which LNL and MU baselines are adapted to the MRR setting is not specified, and several reported baseline numbers are implausible under any standard LNL protocol. The text only says that hyperparameters were 'carefully tuned' and gives no initialization rule (from f(θ0), from f(θu), or from scratch), no training data subset (D_u, D_c^u, or D_n^u), and no epoch or stopping criterion. At CIFAR-10 with 10% noise, Co-teaching is reported at 45.12%, a result far below the Degrade reference of 84.82%; at CIFAR-100 with 10% noise, Co-teaching is 37.75% versus Degrade at 63.70%. On a 90%-clean dataset Co-teaching should be much closer to standard noisy-label training performance, so these numbers suggest an artifact in the adaptation rather than genuine method behavior. Because every claim of 'consistently outperforms' is built on these comparisons, the paper must state the exact adaptation protocol for every baseline and reconcile or correct these anomalous entries.
  2. [Section 3.3, Eqs. (6)-(12), Algorithm 1] The object being unlearned is ambiguous. In Eq. (6), Y^u is defined as the student's predictive label, and Eq. (9) builds S^u_τ from that predicted label. Eq. (11) then applies label smoothing to y^u and Eq. (12) runs gradient ascent on that quantity. The surrounding prose nevertheless describes S^u_τ as 'high-confidence noisy label data,' which suggests that the observed noisy labels from D_u should be the unlearning targets. If the implementation unlearns the student's own predicted labels, the method is not unlearning the influence of the noisy labels but of the model's current beliefs on disagreement samples; if it unlearns the observed labels, Eqs. (6)-(10) use a misleading notation. This distinction changes the mechanism and must be clarified. The same section also contains a concrete mismatch: Algorithm 1 defines A^u_τ and A^t_τ using c(x; θ_t, θ_u), whereas Eq. (14) defines the agreement confidence as c(x; θ^ul_u, θ_t) using the unlearned student.
  3. [Section 4.1, Tables 3 and 6] No experiment is run with multiple seeds, and no error bars, standard deviations, or significance tests are reported. This matters because several claimed wins are small: at 10% noise on Oxford-IIIT Pet, COLUR (92.59%) leads PENCIL (92.01%) by 0.58 points, and on CIFAR-100, COLUR (66.78%) leads PENCIL (65.02%) by 1.76 points. With a single run and no protocol details it is impossible to tell whether these margins are real or within run-to-run variation. At minimum, the paper should report means and standard deviations over at least three seeds for all methods and noise levels.
  4. [Section 9, Algorithm 1, Table 5] The number of unlearning/relearning iterations N is never reported, and the unlearning update in Eq. (12) has no associated epoch count or stopping criterion. Table 5 lists τ_c, α_mix, α_ls, learning rates, and batch size, but not N, despite Algorithm 1 stating that unlearning and relearning are 'alternatively executed for N iterations.' Since the ablation shows that the unlearning module contributes several accuracy points, N is a load-bearing hyperparameter; the manuscript must report it per dataset and noise level or state the convergence criterion used.
minor comments (7)
  1. [Algorithm 1, lines 18-19 and Eqs. (19), (21)] The lines labeled 'Relearn student model' in fact update θ_t, the teacher; correct the labels.
  2. [Eq. (17)] The Mixup operation mixes SA_mix_<τ with A_mix_τ, but no pairing rule is given for samples from the two sets; specify how the sets are matched or subsampled.
  3. [Figure 7 caption] The caption says CIFAR-100 under 'Symmetric Label Noise Ratios,' but Section 7.1 and Table 2 say CIFAR-100 uses asymmetric noise; fix the caption.
  4. [Algorithm 1] The definition of A^u_τ uses 'for (x, y_u, p_u) ∈ S^u' where it should be '∈ A^u'; fix the typo.
  5. [Section 4.1, Training Details] The text says AdamW and SGD were tried, but does not state which optimizer was finally used for each model; report the final configuration.
  6. [Eqs. (15)-(16)] Eq. (15) uses β_m while Eq. (16) uses β without a subscript; unify the Mixup notation.
  7. [Figure 4] The figure would be easier to interpret with axis labels, a color scale, and panel annotations; as printed, the reader cannot tell what the matrices display besides class indices.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: COLUR's held-out restoration results are measured independently of the selection and relearning machinery, and the cited self-works are motivational rather than load-bearing.

full rationale

The paper's central claim is that COLUR restores a degraded model to high held-out accuracy on CIFAR-10, CIFAR-100, Flower-102, and Oxford-IIIT Pet. These accuracies are computed on the test split D_ts, which is disjoint from the noisy training data D_u used for unlearning and relearning. Nothing in the derivation fits parameters to D_ts, and no equation equates the reported test accuracy to the teacher/student confidence or soft-label construction. The unlearning target set is defined by joint confidence in Eqs. (8)-(10), and the relearning labels are averages or mixups of teacher and student predictions in Eqs. (14)-(21); this is a proposed mechanism, not a tautological restatement of the evaluation metric. The teacher model f(theta_0) is an input to the pipeline, not a quantity derived from the outputs, so using it as a reference reflects an assumption about the original model's reliability rather than a circular reduction. The self-citations, such as Gong et al. 2024, Bao et al. 2025, and Miao et al. 2024, appear only in the introductory 'learning, unlearning and relearning' inspiration and are not used to justify the empirical superiority or to define the method's equations. The auxiliary in-sample 'error labeling rate on D_u_n' metric is an evaluation of label correction on the training noisy subset, but it is not the basis for the central held-out restoration claim, and even if it were contestable, that would be an experimental-validity concern rather than a derivation-level circularity. Concerns about unspecified baseline adaptation protocols are correctness risks, not evidence of circularity, since the paper's own formula for COLUR does not reduce to any baseline's output. Therefore no circular step can be exhibited from the text.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The claim rests on six hand-chosen hyperparameters (τ_c, α_mix, α_ls, λ_u, λ_t, and the unreported N) and five domain assumptions about teacher reliability, confident-disagreement-as-noise, gradient-ascent unlearning safety, mixup benefits, and recoverability. No invented entities are introduced. The central empirical comparison is not derived from these parameters in a circular way; rather, the parameters are fixed operating points whose sensitivity is never examined, so the main risk is that reported margins depend on favorable settings.

free parameters (6)
  • τ_c (confidence threshold) = 0.75
    Fixed at 0.75 for every dataset and noise ratio in Table 5; controls which disagreements are unlearned (Eqs. (9)-(10)) and which agreements are relearned (Eq. (14)). No sensitivity analysis is reported, so the central result depends on this hand-chosen operating point.
  • α_mix (mixup Beta parameter) = 0.75
    Fixed at 0.75 in Table 5 for Eqs. (15)-(16); shapes how low-confidence soft labels are blended with high-confidence ones when building the relearning set.
  • α_ls (label smoothing rate) = 0.25
    Fixed at 0.25 in Table 5 for relearning on agreement data (Eqs. (20)-(21)) and used for the unlearning smoothing (Eq. (11), γ).
  • λ_u (student learning rate) = 2e-4 to 1e-3 per condition
    Varies with dataset and noise level (Table 5), with the text stating a smaller λ_u is preferred at 10% noise and larger at 90%; this per-condition tuning knob affects the reported accuracies.
  • λ_t (teacher learning rate) = 1e-4
    Fixed at 1e-4 across all settings (Table 5) for the teacher-update equations (19) and (21).
  • N (unlearning/relearning iterations) = not reported
    Algorithm 1 says the unlearning and relearning steps 'are alternatively executed for N iterations', but N never appears in Table 5 or the text; the number of alternations is an unstated tunable that the final accuracy depends on.
assumptions (5)
  • domain assumption The pre-degradation model f(θ_0) provides reliable high-confidence predictions on the new, possibly noisy dataset D_u.
    The teacher is initialized as a copy of θ_0 (Section 3.3, Eq. (6)) and its confident predictions drive both unlearning selection (Eqs. (8)-(10)) and relearning targets (Eq. (14)). If θ_0 is weak on D_u, the selection is poisoned.
  • domain assumption A sample where teacher and student both predict different labels with high joint confidence is very likely mislabeled.
    This is the selection heuristic underlying S_u_τ (Eqs. (8)-(10)); the paper offers intuition but no analysis or reference validating the heuristic on ambiguous (implicit) noise.
  • domain assumption Label-smoothed gradient ascent on the selected subset removes the influence of noisy labels without damaging the student's useful knowledge.
    The unlearning step (Eqs. (11)-(12)) is borrowed from Di et al. 2024; the paper does not verify unlearning quality via removal or membership tests, and the ablation shows UL alone adds only 0.78 points on Flower-102.
  • domain assumption Mixup between low-confidence soft labels and high-confidence soft labels yields training targets that mitigate label noise.
    The relearning set is built with mixup (Eqs. (15)-(17)) following Carratino et al. 2022 and Zhang et al. 2018; the benefit is assumed to transfer to the MRR setting.
  • domain assumption Performance degradation from incremental training on noisy D_u is recoverable without access to the original training data D_0.
    This is the MRR problem premise (Eq. (3) and Section 3.2); the entire framework is built on the recoverability assumption, tested only indirectly via the reported accuracies.

how reviews work

0 comments
Cite this review

Pith. "Pith review of COLUR: Confidence-Oriented Learning, Unlearning and Relearning with Noisy-Label Data for Model Restoration and Refinement." pith.science (2026). https://pith.science/paper/GCFBDQKV

@misc{pith2026250619496,
  author       = {Pith},
  title        = {Pith review of: COLUR: Confidence-Oriented Learning, Unlearning and Relearning with Noisy-Label Data for Model Restoration and Refinement},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GCFBDQKV}},
  note         = {Machine review of arXiv:2506.19496}
}
read the original abstract

Large deep learning models have achieved significant success in various tasks. However, the performance of a model can significantly degrade if it is needed to train on datasets with noisy labels with misleading or ambiguous information. To date, there are limited investigations on how to restore performance when model degradation has been incurred by noisy label data. Inspired by the ``forgetting mechanism'' in neuroscience, which enables accelerating the relearning of correct knowledge by unlearning the wrong knowledge, we propose a robust model restoration and refinement (MRR) framework COLUR, namely Confidence-Oriented Learning, Unlearning and Relearning. Specifically, we implement COLUR with an efficient co-training architecture to unlearn the influence of label noise, and then refine model confidence on each label for relearning. Extensive experiments are conducted on four real datasets and all evaluation results show that COLUR consistently outperforms other SOTA methods after MRR.

Figures

Figures reproduced from arXiv: 2506.19496 by the authors.

Figure 1
Figure 1. An illustration of labeling on two classes: cat and dog. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. (a) Inspiration of LUR and (b) Scheme of COLUR Scheme for MRR. (1) Learning: The model f(θ 0 ) is incrementally trained on a noisy label dataset leading to a degraded model f(θ u ); (2) Unlearning: MU is employed to unlearn high-confidence noisy label data, resulting in model f(θ u ul); (3) Relearning: the confidence-refined dataset is constructed for relearning f(θ u ul), finally leading to the refined model f(θ u … view at source ↗
Figure 3
Figure 3. The Accuracies After Label Correction. Each radar chart compares the accuracies after the model restoration on 25% and 75% noise-ratio datasets: CIFAR-10, CIFAR-100, Flower-102, and Oxford-IIIT Pet. The Gray-shaded area represents each method’s performance, with a larger area indicating better noise-handling capability. Our COLUR method, outlined in Orange, consistently covers a broader area compared to LNL and MU m… view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Confusion Matrices under Very Low and High Noise Levels. The top two rows display the confusion matrices for CIFAR￾10, while the bottom two rows for Oxford-IIIT Pet. Each matrix compares the performance of various methods at low (10%) and high (90%) label noise ratios.…
Figure 5
Figure 5. Figure 5: Error Rates After Label Refinement. Classification error rates on the noisy label subset D u n of CIFAR-10, CIFAR-100, Flower￾102, and Oxford-IIIT Pet datasets with 50% noise ratio. Degrade shows degradation on noisy data, while LNL methods, MU methods, and COLUR (high…
Figure 6
Figure 6. Figure 6: Class-wise Sample Distributions for CIFAR-10 under Symmetric Label Noise Ratios η ∈ {0.1, 0.25, 0.5, 0.75, 0.9}. Each subfigure illustrates the number of Original (blue), Noisy (orange), and Clean (green) samples for 10 classes. At low noise ratios (η = 0.1 and 0.25), …
Figure 7
Figure 7. Figure 7: Class-wise Sample Distributions for CIFAR-100 under Symmetric Label Noise Ratios η ∈ {0.1, 0.25, 0.5, 0.75, 0.9}. Each subfigure illustrates the number of Original (blue), Noisy (orange), and Clean (green) samples across 100 classes. At low noise ratios (η = 0.1 and 0.…
Figure 8
Figure 8. Figure 8: Error Rates After Label Correction. Classification error rates on the noisy label subset D u n of CIFAR-10, CIFAR-100, Flower￾102, and Oxford-IIIT Pet datasts with 10%/25%/75%/90% noise ratios. Degrade shows degradation on noisy data, while LNL methods, MU methods, and…
Figure 9
Figure 9. Figure 9: The Accuracies After Label Correction. Each radar chart compares the accuracies after the model restoration on 10%, 50%, and 90% noise-ratio datasets: CIFAR-10, CIFAR-100, Flower-102, and Oxford-IIIT Pet. The Gray-shaded area represents each method’s performance, with …
Figure 10
Figure 10. Figure 10: t-SNE Visualization of Feature Representations on CIFAR-10 Dataset. This figure presents the class discrimination capabilities on the CIFAR-10 dataset using t-SNE to visualize feature representations. For CIFAR-10, we added 10%/25%/50%75%/90% symmetric noise. Each col…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

55 extracted references · 49 canonical work pages

  1. [1]

    Image classification with deep learning in the presence of noisy labels: A survey

    G \"o rkem Algan and Ilkay Ulusoy. Image classification with deep learning in the presence of noisy labels: A survey. Knowledge-Based Systems , 215:106771, 2021

  2. [2]

    Wills aligner: Multi-subject collaborative brain visual decoding

    Guangyin Bao, Qi Zhang, Zixuan Gong, Jialei Zhou, Wei Fan, Kun Yi, Usman Naseem, Liang Hu, and Duoqian Miao. Wills aligner: Multi-subject collaborative brain visual decoding. In AAAI , pages 14194--14202. AAAI Press, 2025

  3. [3]

    Evaluating machine unlearning via epistemic uncertainty

    Alexander Becker and Thomas Liebig. Evaluating machine unlearning via epistemic uncertainty. ArXiv preprint arXiv:2208.10836 , 2022

  4. [4]

    On mixup regularization

    Luigi Carratino, Moustapha Ciss \'e , Rodolphe Jenatton, and Jean-Philippe Vert. On mixup regularization. Journal of Machine Learning Research , 23(325):1--31, 2022

  5. [5]

    Machine unlearning via null space calibration

    Huiqiang Chen, Tianqing Zhu, Xin Yu, and Wanlei Zhou. Machine unlearning via null space calibration. In Kate Larson, editor, Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence, IJCAI-24 , pages 358--366. International Joint Conferences on Artificial Intelligence Organization, 8 2024. Main Track

  6. [6]

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

    Hao Cheng, Zhaowei Zhu, Xingyu Li, Yifei Gong, Xing Sun, and Yang Liu. Learning with instance-dependent label noise: A sample sieve approach. In ICLR , 2021

  7. [7]

    Label smoothing improves machine unlearning, 2024

    Zonglin Di, Zhaowei Zhu, Jinghan Jia, Jiancheng Liu, Zafar Takhirov, Bo Jiang, Yuanshun Yao, Sijia Liu, and Yang Liu. Label smoothing improves machine unlearning, 2024

  8. [8]

    Learning, unlearning, and relearning: Using web 2.0 technologies to support the development of lifelong learning skills

    Joanna C Dunlap and Patrick R Lowenthal. Learning, unlearning, and relearning: Using web 2.0 technologies to support the development of lifelong learning skills. In E-infrastructures and technologies for lifelong learning: Next generation environments , pages 292--315. IGI Global, 2011

Show all 55 references
  1. [9]

    Generalized jensen-shannon divergence loss for learning with noisy labels

    Erik Englesson and Hossein Azizpour. Generalized jensen-shannon divergence loss for learning with noisy labels. NeurIPS , 34:30284--30297, 2021

  2. [10]

    Salun: Empowering machine unlearning via gradient-based weight saliency in both image classification and generation

    Chongyu Fan, Jiancheng Liu, Yihua Zhang, Eric Wong, Dennis Wei, and Sijia Liu. Salun: Empowering machine unlearning via gradient-based weight saliency in both image classification and generation. In ICLR , 2024

  3. [11]

    Aritra Ghosh, Himanshu Kumar, and P. S. Sastry. Robust loss functions under label noise for deep neural networks. In AAAI , 2017

  4. [12]

    Eternal sunshine of the spotless net: Selective forgetting in deep networks

    Aditya Golatkar, Alessandro Achille, and Stefano Soatto. Eternal sunshine of the spotless net: Selective forgetting in deep networks. In CVPR , pages 9304--9312, 2020

  5. [13]

    Neuroclips: Towards high-fidelity and smooth fmri-to-video reconstruction

    Zixuan Gong, Guangyin Bao, Qi Zhang, Zhongwei Wan, Duoqian Miao, Shoujin Wang, Lei Zhu, Changwei Wang, Rongtao Xu, Liang Hu, Ke Liu, and Yu Zhang. Neuroclips: Towards high-fidelity and smooth fmri-to-video reconstruction. In NeurIPS , 2024

  6. [14]

    Amnesiac machine learning

    Laura Graves, Vineel Nagisetty, and Vijay Ganesh. Amnesiac machine learning. In Proceedings of the AAAI Conference on Artificial Intelligence , volume 35, pages 11516--11524, 2021

  7. [15]

    Tsang, and Masashi Sugiyama

    Bo Han, Quanming Yao, Xingrui Yu, Gang Niu, Miao Xu, Weihua Hu, Ivor W. Tsang, and Masashi Sugiyama. Co-teaching: Robust training of deep neural networks with extremely noisy labels. In NeurIPS , 2018

  8. [16]

    Approximate data deletion from machine learning models

    Zachary Izzo, Mary Anne Smart, Kamalika Chaudhuri, and James Zou. Approximate data deletion from machine learning models. In International Conference on Artificial Intelligence and Statistics , pages 2008--2016. PMLR, 2021

  9. [17]

    Mentornet: Learning data-driven curriculum for very deep neural networks on corrupted labels

    Lu Jiang, Zhengyuan Zhou, Thomas Leung, Li - Jia Li, and Li Fei - Fei. Mentornet: Learning data-driven curriculum for very deep neural networks on corrupted labels. In ICML , volume 80, 2018

  10. [18]

    UNICON: combating label noise through uniform selection and contrastive learning

    Nazmul Karim, Mamshad Nayeem Rizve, Nazanin Rahnavard, Ajmal Mian, and Mubarak Shah. UNICON: combating label noise through uniform selection and contrastive learning. In CVPR , 2022

  11. [19]

    Nlnl: Negative learning for noisy labels

    Youngdong Kim, Junho Yim, Juseung Yun, and Junmo Kim. Nlnl: Negative learning for noisy labels. In CVPR , pages 101--110, 2019

  12. [20]

    Overcoming catastrophic forgetting in neural networks

    James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, et al. Overcoming catastrophic forgetting in neural networks. PNAS , 114(13):3521--3526, 2017

  13. [21]

    Learning, unlearning, and relearning: Lessons from one school's approach to creating and sustaining learning communities

    Emily J Klein. Learning, unlearning, and relearning: Lessons from one school's approach to creating and sustaining learning communities. Teacher Education Quarterly , 35(1):79--97, 2008

  14. [22]

    Understanding black-box predictions via influence functions

    Pang Wei Koh and Percy Liang. Understanding black-box predictions via influence functions. In ICML , pages 1885--1894. PMLR, 2017

  15. [23]

    Learning multiple layers of features from tiny images

    Alex Krizhevsky, Geoffrey Hinton, et al. Learning multiple layers of features from tiny images. Master’s Thesis , 2009

  16. [24]

    Junnan Li, Richard Socher, and Steven C. H. Hoi. Dividemix: Learning with noisy labels as semi-supervised learning. In ICLR , 2020

  17. [25]

    Junnan Li, Caiming Xiong, and Steven C. H. Hoi. Mopro: Webly supervised learning with momentum prototypes. In ICLR , 2021

  18. [26]

    Disc: Learning from noisy labels via dynamic instance-specific selection and correction

    Yifan Li, Hu Han, Shiguang Shan, and Xilin Chen. Disc: Learning from noisy labels via dynamic instance-specific selection and correction. In CVPR , pages 24070--24079, 2023

  19. [27]

    Early-learning regularization prevents memorization of noisy labels

    Sheng Liu, Jonathan Niles - Weed, Narges Razavian, and Carlos Fernandez - Granda. Early-learning regularization prevents memorization of noisy labels. In NeurIPS , 2020

  20. [28]

    Model sparsity can simplify machine unlearning

    Jiancheng Liu, Parikshit Ram, Yuguang Yao, Gaowen Liu, Yang Liu, PRANAY SHARMA, Sijia Liu, et al. Model sparsity can simplify machine unlearning. NeurIPS , 36, 2024

  21. [29]

    Does label smoothing mitigate label noise? In ICML , pages 6448--6458

    Michal Lukasik, Srinadh Bhojanapalli, Aditya Menon, and Sanjiv Kumar. Does label smoothing mitigate label noise? In ICML , pages 6448--6458. PMLR, 2020

  22. [30]

    when to update

    Eran Malach and Shai Shalev-Shwartz. Decoupling" when to update" from" how to update". NeurIPS , 30, 2017

  23. [31]

    Graph memory learning: Imitating lifelong remembering and forgetting of brain networks

    Jiaxing Miao, Liang Hu, Qi Zhang, and Longbing Cao. Graph memory learning: Imitating lifelong remembering and forgetting of brain networks. CoRR , abs/2407.19183, 2024

  24. [32]

    Automated flower classification over a large number of classes

    Maria - Elena Nilsback and Andrew Zisserman. Automated flower classification over a large number of classes. In ICVGIP , 2008

  25. [33]

    O'Connor, and Kevin McGuinness

    Diego Ortego, Eric Arazo, Paul Albert, Noel E. O'Connor, and Kevin McGuinness. Multi-objective interpolation training for robustness to label noise. In CVPR , 2021

  26. [34]

    Parkhi, Andrea Vedaldi, Andrew Zisserman, and C

    Omkar M. Parkhi, Andrea Vedaldi, Andrew Zisserman, and C. V. Jawahar. Cats and dogs. In CVPR , 2012

  27. [35]

    Learn, unlearn and relearn: An online learning paradigm for deep neural networks

    Vijaya Raghavan T Ramkumar, Elahe Arani, and Bahram Zonooz. Learn, unlearn and relearn: An online learning paradigm for deep neural networks. Transactions on Machine Learning Research , 2023

  28. [36]

    Forgetting as a form of adaptive engram cell plasticity

    Tom \'a s J Ryan and Paul W Frankland. Forgetting as a form of adaptive engram cell plasticity. Nature Reviews Neuroscience , 23(3):173--186, 2022

  29. [37]

    Noisy concurrent training for efficient learning under label noise

    Fahad Sarfraz, Elahe Arani, and Bahram Zonooz. Noisy concurrent training for efficient learning under label noise. In Proceedings of the IEEE/CVF Winter Conference on applications of computer vision , pages 3159--3168, 2021

  30. [38]

    forgetting

    Alyssa Shuang Sha, Bernardo Pereira Nunes, and Armin Haller. " forgetting" in machine learning and beyond: A survey. arXiv preprint arXiv:2405.20620 , 2024

  31. [39]

    Rethinking the inception architecture for computer vision

    Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jon Shlens, and Zbigniew Wojna. Rethinking the inception architecture for computer vision. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 2818--2826, 2016

  32. [40]

    Efficientnet: Rethinking model scaling for convolutional neural networks

    Mingxing Tan and Quoc Le. Efficientnet: Rethinking model scaling for convolutional neural networks. In ICML , pages 6105--6114. PMLR, 2019

  33. [41]

    Symmetric cross entropy for robust learning with noisy labels

    Yisen Wang, Xingjun Ma, Zaiyi Chen, Yuan Luo, Jinfeng Yi, and James Bailey. Symmetric cross entropy for robust learning with noisy labels. In ICCV , 2019

  34. [42]

    Machine unlearning of features and labels

    Alexander Warnecke, Lukas Pirch, Christian Wressnegger, and Konrad Rieck. Machine unlearning of features and labels. arXiv preprint arXiv:2108.11577 , 2021

  35. [43]

    Combating noisy labels by agreement: A joint training method with co-regularization

    Hongxin Wei, Lei Feng, Xiangyu Chen, and Bo An. Combating noisy labels by agreement: A joint training method with co-regularization. In CVPR , pages 13726--13735, 2020

  36. [44]

    To smooth or not? when label smoothing meets noisy labels

    Jiaheng Wei, Hangyu Liu, Tongliang Liu, Gang Niu, and Yang Liu. To smooth or not? when label smoothing meets noisy labels. In ICML , 2022

  37. [45]

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

    Yilun Xu, Peng Cao, Yuqing Kong, and Yizhou Wang. L \_ dmi: A novel information-theoretic loss function for training deep nets robust to label noise. In NeurIPS , 2019

  38. [46]

    Heng Xu, Tianqing Zhu, Lefeng Zhang, Wanlei Zhou, and Philip S. Yu. Machine unlearning: A survey. ACM Comput. Surv. , 56(1), August 2023

  39. [47]

    Probabilistic end-to-end noise correction for learning with noisy labels

    Kun Yi and Jianxin Wu. Probabilistic end-to-end noise correction for learning with noisy labels. In CVPR , pages 7017--7025, 2019

  40. [48]

    Tsang, and Masashi Sugiyama

    Xingrui Yu, Bo Han, Jiangchao Yao, Gang Niu, Ivor W. Tsang, and Masashi Sugiyama. How does disagreement help generalization against label corruption? In ICML , volume 97, 2019

  41. [49]

    Wide residual networks

    Sergey Zagoruyko and Nikos Komodakis. Wide residual networks. In BMVC , 2016

  42. [50]

    Zhilu Zhang and Mert R. Sabuncu. Generalized cross entropy loss for training deep neural networks with noisy labels. In NeurIPS , 2018

  43. [51]

    Dauphin, and David Lopez-Paz

    Hongyi Zhang, Moustapha Cisse, Yann N. Dauphin, and David Lopez-Paz. mixup: Beyond empirical risk minimization. International Conference on Learning Representations , 2018

  44. [52]

    Tripartite collaborative filtering with observability and selection for debiasing rating estimation on missing-not-at-random data

    Qi Zhang, Longbing Cao, Chongyang Shi, and Liang Hu. Tripartite collaborative filtering with observability and selection for debiasing rating estimation on missing-not-at-random data. In AAAI , pages 4671--4678. AAAI Press, 2021

  45. [53]

    Learning with feature-dependent label noise: A progressive approach

    Yikai Zhang, Songzhu Zheng, Pengxiang Wu, Mayank Goswami, and Chao Chen. Learning with feature-dependent label noise: A progressive approach. In ICLR , 2021

  46. [54]

    Learning with noisy labels via sparse regularization

    Xiong Zhou, Xianming Liu, Chenyang Wang, Deming Zhai, Junjun Jiang, and Xiangyang Ji. Learning with noisy labels via sparse regularization. In CVPR , pages 72--81, 2021

  47. [55]

    write newline

    " write newline "" before.all 'output.state := FUNCTION fin.entry add.period write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence '...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.