Pith. sign in

REVIEW 5 major objections 6 minor 88 references

Enhancing Sample Utilization in Noise-Robust Deep Metric Learning With Subgroup-Based Positive-Pair Selection

T0 review · 5 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A noise-robust metric learning framework reuses noisy samples by building reliable positive pairs from subgroup labels, and reports state-of-the-art retrieval and face recognition results.

desk verdict A useful noisy-DML method with a strong face-recognition result, but the paper never measures the precision of the positive pairs it constructs for noisy samples—ask for that diagnostic. read the letter →

arxiv 2501.11063 v1 pith:ZXRBHVRW submitted 2025-01-19 cs.CV

Zhipeng Yu , Qianqian Xu , Yangbangyan Jiang , Yingfei Sun , Qingming Huang This is my paper · ORCID
classification cs.CV
keywords noise-robustdeepmetriclearningpositive-pairselectionsubgroupgenerationlabelnoisesampleutilizationcontrastiveimageretrievalfacerecognition
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 argues that in deep metric learning, noisy labels should not be discarded: they can be turned into useful training signal. It proposes SGPS, which first splits each batch into clean and noisy samples, then discovers likely true positive partners for each noisy sample from subgroup labels derived from the feature distribution, aggregates those partners into a prototype, and trains the noisy sample against that prototype with a contrastive loss. The claim is that this improves sample utilization and outperforms prior noise-robust metric learning methods, which mostly throw noisy samples away. The strongest reported evidence is on face recognition: DCQ+SGPS trained on the noisy MS1MV0 dataset reaches 92.91% TAR@FAR=1e-5 on IJB-C, surpassing DCQ trained on the clean MS1MV2 dataset (91.99%).

What carries the argument

The central object is the subgroup-based positive-pair selection pipeline inside SGPS. In it, a probability-based clean-sample selection (PCS) uses a memory bank of historical features to compute pclean(i) (Eq. 1) and splits each batch into Bclean and Bnoise. The Subgroup Generation Module (SGM) produces two subgroup labelings: bottom-up labels cB from agglomerative merging of intra-class connected components (Algorithms 1-2) and complementary top-down labels cT from recursive maximum-margin hyperplane division (Algorithms 3-4). The Positive Prototype Generation Module (PPM) selects K neighbors of each noisy sample according to cB and cT, and aggregates them into a prototype ri either by mean, max, softmax correlation weights, or a learnable cross-attention transformer (TransProto). The load-bearing step is Eq. (13)/(15): the noisy contrastive loss Lnoise = γ1 Lnoise_batch + γ2 Lnoise_bank pulls each noisy sample toward its prototype and away from negatives whose annotated label and both subgroup labels differ, which is how discarded samples become usable.

What would settle it

Train SGPS and PRISM on CARS with 50% symmetric noise, where ground-truth labels are known, and measure the precision of the K positive samples SGM selects for each noisy sample (fraction whose true label matches). If that precision is at or below the base rate while SGPS still beats PRISM, the subgroup discovery is not the source of the gain. A direct control: replace the subgroup labels with a random partition of each annotated class into the same number of subgroups; if P@1 stays the same, the subgroup machinery is not load-bearing.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that the positive pairs needed for noisy samples can be inferred from the structure of the learned feature space, without trusting the noisy labels. The Subgroup Generation Module (SGM) splits each annotated class into connected components of high intra-class similarity, merges these subgroups bottom-up by centroid similarity under meta-cluster and size rules, and adds a complementary top-down recursive partitioning; the resulting subgroup labels cB and cT mark, for each noisy sample, a set of samples likely to share the true class. The Positive Prototype Generation Module (PPM) then pools K such samples into a single prototype ri, and a new contrastive loss Lnoise pulls the noisy sample toward ri while pushing it away from samples outside its subgroups. Because both clean and noisy samples contribute to training, the framework avoids the confirmation-bias overfitting that the paper documents in the discard-based PRISM method, and across CARS, CUB, SOP, Food-101N, Clothing1M, and MS1MV0 it outperforms the compared noisy-label DML baselines.

Load-bearing premise

The load-bearing premise is that the current feature embedding already forms semantically coherent clusters, so that the connected components in Eq. (4) and the merged centroids of Algorithms 2-3 recover true positive pairs for noisy samples; early in training or at 90% noise this is unverified, and if the clusters are wrong, Lnoise pulls noisy samples toward incorrect prototypes and could amplify the very noise it is meant to remove.

Editorial extensions

If this is right

  • Noisy samples contribute to learning instead of being filtered out, so models can continue improving late in training where discard-based methods like PRISM show overfitting (Fig. 8).
  • SGPS is a wrapper: instantiating it on MCL, SupCon, Roadmap, and Contextual improves all four baselines on synthetic and real-world noisy retrieval benchmarks.
  • On large-scale face recognition, DCQ+SGPS trained on noisy MS1MV0 reaches 92.91% TAR@FAR=1e-5 on IJB-C, surpassing DCQ trained on the clean MS1MV2 (91.99%), implying the noisy data's extra volume can outweigh its label noise.
  • Training cost stays manageable: SGPS runs about 1.1-1.2x slower than PRISM on CARS and SOP and adds no GPU memory overhead, unlike multi-proxy classifier methods (Tab. VI, Fig. 9).

Reading between the lines

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

  • If subgroup labels reliably identify true positives for noisy samples, the same mechanism could refine negative-pair mining (the paper already excludes same-subgroup negatives in Eq. 14) and could be applied to semi-supervised pseudo-labeling, where the noisy set is treated as unlabeled data.
  • The face-recognition result suggests a data-engineering consequence the paper does not state: web-scale noisy collections like MS1MV0 could substitute for expensively cleaned versions when the noise is clusterable, potentially changing dataset curation practice.
  • A testable diagnostic follows from the paper's own assumption: track the precision of SGM-selected positive pairs against ground truth on synthetic noise; if SGPS's advantage over PRISM does not track that precision, the gain may come from the prototype loss rather than subgroup discovery.
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

5 major / 6 minor

Summary. The paper proposes SGPS, a noise-robust deep metric learning framework that, instead of discarding suspicious noisy samples, attempts to construct positive pairs for them from subgroup labels. A probability-based clean-sample selection (PCS) splits each batch into clean and noisy sets; a subgroup generation module (SGM) produces bottom-up and top-down subgroup labels from a momentum feature bank; a positive prototype generation module (PPM) aggregates K selected positives into a prototype per noisy sample; and a contrastive loss L_noise pulls noisy samples toward those prototypes while excluding same-subgroup negatives. The clean set is trained with a standard DML loss L_clean. Experiments cover CARS, SOP, CUB, Food-101N, CARS-98N, Clothing1M, and MS1MV0 face recognition, with SGPS instantiated on MCL, SupCon, Roadmap, and Contextual. The headline result is DCQ+SGPS reaching 92.91% TAR@FAR=1e-5 on IJB-C when trained on noisy MS1MV0, surpassing DCQ trained on clean MS1MV2 (91.99%).

Significance. If the central claim holds, the paper addresses a real gap: prior noise-robust DML methods such as PRISM discard noisy samples, and SGPS offers a general wrapper that can be plugged into several pairwise DML losses. The experimental scope is broad, spanning synthetic symmetric noise, small-cluster noise, real-world noisy image retrieval, and large-scale face recognition, and the framework shows consistent improvements over its base losses in most configurations. The code is promised publicly, which supports reproducibility if the hyperparameters are fully specified. However, the paper's central effectiveness claim is not yet statistically supported: results are single runs without error bars, many SGM hyperparameters are unspecified, and the proposed mechanism's key ingredient is never directly measured. The evidence is broad but does not isolate the contribution of L_noise from confounds such as extra training data or the L_clean branch, so the current support for the abstract's claim of constructing reliable positive pairs for noisy samples is incomplete.

major comments (5)
  1. [IV-C, Table IV] The headline comparison does not isolate the effect of the proposed L_noise branch: DCQ+SGPS on MS1MV0 uses 10M noisy images, while DCQ on clean MS1MV2 uses 5.8M images, and the SGPS training also includes L_clean on selected clean samples. Table IV does not report a control of DCQ on MS1MV0 with only clean-sample selection and L_clean, or a DCQ+SGPS variant with L_noise ablated. Without such controls, the stated conclusion that the gain comes from reliable positive pairs for noisy samples is not identified by the evidence; the gain could come from the larger data volume or from the L_clean branch.
  2. [III-E, III-F, Fig. 6] The manuscript never measures the reliability of the positive pairs Pi that are the load-bearing component of L_noise. Figure 6 reports only clean-sample selection accuracy, not the precision of the K selected positives for noisy anchors, e.g., the fraction of Pi sharing the ground-truth label with the noisy anchor and how this evolves over training or at high noise rates. Because Eq. (13) pulls the noisy sample toward a prototype aggregated from Pi and Eq. (14) excludes same-subgroup samples from negatives by construction, incorrect subgroup labels would cause L_noise to reinforce false positives. The claimed mechanism therefore needs a direct measurement; with only end-task numbers, the alternative explanations in the previous comment remain open.
  3. [III-D, Algorithms 1-4, IV-B] Several SGM hyperparameters that directly control the subgroup labels are not reported for any experiment, including lambda_min, lambda_max, lambda'_min, lambda'_max, tau_max, B, the window omega and percentile R in Eq. (2), the momentum alpha in Eq. (3), and the number K of selected positives. Table V varies only tau, delta, gamma1, and gamma2 on one dataset, and Table VI uses K=4 only for runtime. Without the SGM settings, the experimental results are not reproducible and the sensitivity of the central mechanism to these choices is unknown.
  4. [IV-E, Table I] The 'posterior data clean and training strategy' SGPS-MCL* is a three-stage variant that trains on pseudo-labels, applies an early-stopping PRISM, and uses a stage-3 fine-tune with original labels; Table I presents it as a main result without isolating its contribution from the core SGPS framework. The description leaves unspecified how cB pseudo-labels are generated for stage 2, what noise rate is used in the early-stopping PRISM stage, and which model-selection criterion is applied. Presenting this post-hoc variant in the headline table with the asterisk is acceptable only if its hyperparameters are fully specified and its difference from the base SGPS is clear.
  5. [Tables I-IV] All reported numbers appear to be single runs, with no standard deviations, number of trials, or significance tests. Many of the claimed improvements over the strongest baselines are small relative to likely run-to-run variation, for example Table III CARS-98N P@1 of 73.56 for SGPS-Roadmap versus 64.29 for Roadmap is large, but Table III Clothing1M P@1 differences between SGPS variants are around 0.2-0.5 points, and Table II contains several close values. The paper should report multiple seeds or error bars for at least the main comparisons to support the broad claim of outperforming state-of-the-art methods.
minor comments (6)
  1. [IV-A] The sentence 'The evaluation is conducted on seven image retrieval or face recognition benchmark datasets, including:' is duplicated verbatim in the dataset section.
  2. [III-D, Eq. (4)] The role of lambda_max in Eq. (4) is unclear: W_m_ij is set to 1 if S_m_ij > lambda_max, but Algorithm 1 then sets W_m_ij = 0 if S_m_ij < lambda_min; it should be stated whether lambda_min and lambda_max are used jointly and how overlapping conditions are resolved.
  3. [III-D, Eq. (6), Algorithm 3] The hyperplane is written as h = (f_i - f_j)/2, but the division by 2 is irrelevant for the side test f_k^T h >= 0, and the notation would be clearer if h were defined as the normal vector f_i - f_j (or normalized) with the decision boundary at (f_i + f_j)/2.
  4. [IV-E, Fig. 5(a)] The ablation subsection title 'Effectiveness of GSM' contains a typo; it should be 'SGM'.
  5. [IV-D, Table V] The table header says 'CARS19', but the dataset is referred to as CARS elsewhere; please use the consistent dataset name.
  6. [I, Fig. 1(c)] The method name 'PRISIM' in the Fig. 1 caption should be 'PRISM'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: SGPS is a self-training framework evaluated on held-out benchmarks; no claim reduces by construction to its inputs.

full rationale

The paper's derivation chain is self-contained against external benchmarks. The only self-referential element is that SGM (Eqs. 3-6, Algorithms 1-4) derives subgroup labels from the current feature bank, and Lnoise (Eqs. 13-16) then trains the same feature extractor toward prototypes aggregated from those subgroups. This is a standard self-training/bootstrap mechanism, not a definitional circularity: the subgroup labels are not defined in terms of the final retrieval metric, and the claimed improvement (Tables I-IV) is measured on held-out test sets, so the outcome is not forced by construction. The PCS threshold (Eq. 2) and loss weights are hyperparameters, not fitted to test metrics. The paper cites prior work by its own authors (e.g., refs. 10, 18, 29, 41-42, 45, 50-52) only as related work or as components (PRISM [4] and DCQ [80] are external), and no load-bearing premise rests on a self-citation. The concern that the precision of positive pairs Pi for the noisy subset is never directly reported is a correctness/evidence gap, not a circular reduction; likewise, the Table IV comparison confounds data scale with noise handling but does not make the result equivalent to its inputs. No equation reduces to another by construction, and no fitted parameter is renamed as a prediction. Score 0.

Assumptions & free parameters 10 free parameters · 6 assumptions · 4 invented entities

The framework rests on many hand-set thresholds and internal constructs derived from the model's own features. No free parameter is fitted to a held-out test metric, but the lack of a validation protocol for SGM thresholds and the absence of per-seed variance leave the empirical claim only partially secured.

free parameters (10)
  • temperature tau in Lnoise = 0.02 (best on CARS 50% symmetric noise)
    Controls sharpness of contrastive loss; sensitivity table shows varying it changes P@1 by about 3 points.
  • margin delta in Lnoise = 0.1 (best)
    Controls separation between positive prototype and negatives; too large causes overfitting.
  • loss weights gamma1 and gamma2 = 1.0 and 0.1 (default)
    Balance batch and bank losses for noisy samples; sensitivity shows modest effect.
  • K (number of selected positive samples per noisy sample) = 4 (used in runtime experiments)
    Number of positive samples aggregated into prototype; not justified analytically.
  • thresholds lambda_min and lambda_max in intra-class splitting = not specified
    Define adjacency in Eq.4; values affect subgroup purity but no sensitivity analysis is provided.
  • parameters lambda'_min, lambda'_max, tau_k, tau_max in bottom-up merging = not fully specified; tau_k set to twice class count for real-world datasets
    Control merging of subgroups; adjusted per dataset, raising tuning concerns.
  • B (max samples per cell in top-down division) = not specified
    Governs recursive partitioning; no analysis of its effect.
  • window omega and percentile R in clean-sample threshold = not specified
    Control the threshold schedule adopted from PRISM; inherited tuning dependence.
  • momentum alpha for feature bank = not specified
    Sets update speed of features in Eq.3; affects subgroup quality.
  • TransProto transformer hyperparameters = 3 layers, dimensions not given
    Learned prototype aggregator; training details absent.
assumptions (6)
  • domain assumption Samples close in the embedding share ground-truth category, so connected components of the intra-class similarity graph are purer than the noisy class labels.
    Invoked in Algorithm 1 and Eq.4; if false, positive prototypes for noisy samples can be wrong.
  • domain assumption Class centroids w_m computed from the memory bank are reliable, so p_clean in Eq.1 ranks clean samples higher than noisy ones.
    Core to PCS; at high noise or early training, centroids can be corrupted.
  • domain assumption The smooth top-R threshold schedule (Eq.2) with predefined noise ratio R separates clean from noisy samples accurately.
    Adopted from PRISM; assumes the noise fraction is known and stable.
  • domain assumption Momentum-updated feature bank is consistent with the current model despite asynchronous updates.
    Subgroup labels are only refreshed asynchronously; stale features could mislead selection.
  • domain assumption Prototype aggregation (mean/softmax/TransProto) produces a feature closer to the true class center than the noisy sample itself.
    Needed for Lnoise to pull noisy samples toward the true class; not verified externally.
  • standard math Standard connected components labeling and hierarchical merging behave as specified in Algorithms 1-3.
    Routine algorithms; no proof supplied, but not the fragile part.
invented entities (4)
  • Subgroup labels cB and cT
    purpose: Provide candidate positive samples for noisy samples and define negative exclusion sets in Eq.14.
    Generated from the model's own features; no external ground truth.
  • Positive prototype ri
    purpose: Aggregated target for each noisy sample in the contrastive loss Lnoise.
    Computed from selected positive samples; its quality depends on subgroup accuracy.
  • Meta cluster Cm*
    purpose: Largest subgroup per class, used to constrain merging in bottom-up generation.
    Internal construct with no external validation.
  • TransProto module
    purpose: Learned cross-attention aggregator producing prototypes.
    Trained within the pipeline; no independent evidence of superiority beyond one ablation figure.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Enhancing Sample Utilization in Noise-Robust Deep Metric Learning With Subgroup-Based Positive-Pair Selection." pith.science (2026). https://pith.science/paper/ZXRBHVRW

@misc{pith2026250111063,
  author       = {Pith},
  title        = {Pith review of: Enhancing Sample Utilization in Noise-Robust Deep Metric Learning With Subgroup-Based Positive-Pair Selection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZXRBHVRW}},
  note         = {Machine review of arXiv:2501.11063}
}
read the original abstract

The existence of noisy labels in real-world data negatively impacts the performance of deep learning models. Although much research effort has been devoted to improving the robustness towards noisy labels in classification tasks, the problem of noisy labels in deep metric learning (DML) remains under-explored. Existing noisy label learning methods designed for DML mainly discard suspicious noisy samples, resulting in a waste of the training data. To address this issue, we propose a noise-robust DML framework with SubGroup-based Positive-pair Selection (SGPS), which constructs reliable positive pairs for noisy samples to enhance the sample utilization. Specifically, SGPS first effectively identifies clean and noisy samples by a probability-based clean sample selectionstrategy. To further utilize the remaining noisy samples, we discover their potential similar samples based on the subgroup information given by a subgroup generation module and then aggregate them into informative positive prototypes for each noisy sample via a positive prototype generation module. Afterward, a new contrastive loss is tailored for the noisy samples with their selected positive pairs. SGPS can be easily integrated into the training process of existing pair-wise DML tasks, like image retrieval and face recognition. Extensive experiments on multiple synthetic and real-world large-scale label noise datasets demonstrate the effectiveness of our proposed method. Without any bells and whistles, our SGPS framework outperforms the state-of-the-art noisy label DML methods. Code is available at \url{https://github.com/smuelpeng/SGPS-NoiseFreeDML}.

Figures

Figures reproduced from arXiv: 2501.11063 by the authors.

Figure 1
Figure 1. (a) Performance drop corresponding to the noise level for image classification and DML tasks (image retrieval and face recognition). (b) Precision@1 of [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The framework of our proposed method. The input batch will first be fed into the feature extractor network to obtain the features. Then, inputs will be [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The workflow of SGM. SGM will maintain two kinds of subgroup [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Visualization of TransProto PPM. We employ a 3-layer transformer [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 6
Figure 6. Figure 6: Quantitative accuracy of clean sample selection on CARS and SOP [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Qualitative visualization of the detailed selection results in a training [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Change of Precision and MAP during the training. SGPS overcomes [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 10
Figure 10. Figure 10: Visualization of SGM’s results on SOP with 50% symmetric noise. [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

88 extracted references · 73 canonical work pages

  1. [1]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in CVPR, 2016, pp. 770–778. IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 33, 2024 14

    2016

  2. [2]

    Multi- similarity loss with general pair weighting for deep metric learning,

    X. Wang, X. Han, W. Huang, D. Dong, and M. R. Scott, “Multi- similarity loss with general pair weighting for deep metric learning,” in CVPR, 2019, pp. 5022–5030

    2019

  3. [3]

    Arcface: Additive angular margin loss for deep face recognition,

    J. Deng, J. Guo, N. Xue, and S. Zafeiriou, “Arcface: Additive angular margin loss for deep face recognition,” in CVPR, 2019, pp. 4690–4699

    2019

  4. [4]

    Noise-resistant deep metric learning with ranking-based instance selection,

    C. Liu, H. Yu, B. Li, Z. Shen, Z. Gao, P. Ren, X. Xie, L. Cui, and C. Miao, “Noise-resistant deep metric learning with ranking-based instance selection,” in CVPR, 2021, pp. 6811–6820

    2021

  5. [5]

    K-means++ the advantages of careful seeding,

    D. Arthur and S. Vassilvitskii, “K-means++ the advantages of careful seeding,” in SODA, 2007, pp. 1027–1035

    2007

  6. [6]

    Hierarchical clustering,

    F. Nielsen and F. Nielsen, “Hierarchical clustering,” Introduction to HPC with MPI for Data Science , pp. 195–211, 2016

    2016

  7. [7]

    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,” NeurIPS, vol. 31, 2018

    2018

  8. [8]

    Sample selection with uncertainty of losses for learning with noisy labels,

    X. Xia, T. Liu, B. Han, M. Gong, J. Yu, G. Niu, and M. Sugiyama, “Sample selection with uncertainty of losses for learning with noisy labels,” in ICLR, 2022

    2022

Show all 88 references
  1. [9]

    Selective-supervised contrastive learning with noisy labels,

    S. Li, X. Xia, S. Ge, and T. Liu, “Selective-supervised contrastive learning with noisy labels,” in CVPR, 2022, pp. 316–325

    2022

  2. [10]

    Dist-pu: Positive- unlabeled learning from a label distribution perspective,

    Y . Zhao, Q. Xu, Y . Jiang, P. Wen, and Q. Huang, “Dist-pu: Positive- unlabeled learning from a label distribution perspective,” inCVPR, 2022, pp. 14 461–14 470

    2022

  3. [11]

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

    L. Jiang, Z. Zhou, T. Leung, L.-J. Li, and L. Fei-Fei, “Mentornet: Learning data-driven curriculum for very deep neural networks on corrupted labels,” in ICML. PMLR, 2018, pp. 2304–2313

    2018

  4. [12]

    Meta label correction for noisy label learning,

    G. Zheng, A. H. Awadallah, and S. Dumais, “Meta label correction for noisy label learning,” in AAAI, vol. 35, no. 12, 2021, pp. 11 053–11 061

    2021

  5. [13]

    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 CVPR, 2017, pp. 1944–1952

    2017

  6. [14]

    Estimating noise transition matrix with label correlations for noisy multi-label learning,

    “Estimating noise transition matrix with label correlations for noisy multi-label learning,” in 36th Conference on Neural Information Pro- cessing Systems (NeurIPS 2022) , 2022

    2022

  7. [15]

    A parametrical model for instance-dependent label noise,

    S. Yang, S. Wu, E. Yang, B. Han, Y . Liu, M. Xu, G. Niu, and T. Liu, “A parametrical model for instance-dependent label noise,” IEEE TPAMI, vol. 45, no. 12, pp. 14 055–14 068, 2023

    2023

  8. [16]

    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

    2002 arXiv

  9. [17]

    Me-momentum: Extracting hard confident examples from noisily labeled data,

    Y . Bai and T. Liu, “Me-momentum: Extracting hard confident examples from noisily labeled data,” in ICCV, 2021

    2021

  10. [18]

    Maxmatch: Semi-supervised learning with worst-case consistency,

    Y . Jiang, X. Li, Y . Chen, Y . He, Q. Xu, Z. Yang, X. Cao, and Q. Huang, “Maxmatch: Semi-supervised learning with worst-case consistency,” IEEE TPAMI, vol. 45, no. 5, pp. 5970–5987, 2022

    2022

  11. [19]

    Learning a similarity metric discriminatively, with application to face verification,

    S. Chopra, R. Hadsell, and Y . LeCun, “Learning a similarity metric discriminatively, with application to face verification,” in CVPR, vol. 1. IEEE, 2005, pp. 539–546

    2005

  12. [20]

    Cross-batch memory for embedding learning,

    X. Wang, H. Zhang, W. Huang, and M. R. Scott, “Cross-batch memory for embedding learning,” in CVPR, 2020, pp. 6388–6397

    2020

  13. [21]

    Deep image retrieval is not robust to label noise,

    S. Dereka, I. Karpukhin, and S. Kolesnikov, “Deep image retrieval is not robust to label noise,” in CVPR, 2022, pp. 4975–4980

    2022

  14. [22]

    Facenet: A unified embed- ding for face recognition and clustering,

    F. Schroff, D. Kalenichenko, and J. Philbin, “Facenet: A unified embed- ding for face recognition and clustering,” in CVPR, 2015, pp. 815–823

    2015

  15. [23]

    Circle loss: A unified perspective of pair similarity optimization,

    Y . Sun, C. Cheng, Y . Zhang, C. Zhang, L. Zheng, Z. Wang, and Y . Wei, “Circle loss: A unified perspective of pair similarity optimization,” in CVPR, 2020, pp. 6398–6407

    2020

  16. [24]

    Attributable visual similarity learning,

    B. Zhang, W. Zheng, J. Zhou, and J. Lu, “Attributable visual similarity learning,” in CVPR, 2022, pp. 7532–7541

    2022

  17. [25]

    Neighbourhood components analysis,

    J. Goldberger, G. E. Hinton, S. Roweis, and R. R. Salakhutdinov, “Neighbourhood components analysis,” NeurIPS, vol. 17, 2004

    2004

  18. [26]

    Sampling matters in deep embedding learning,

    C.-Y . Wu, R. Manmatha, A. J. Smola, and P. Krahenbuhl, “Sampling matters in deep embedding learning,” in ICCV, 2017, pp. 2840–2848

    2017

  19. [27]

    Deep metric learning to rank,

    F. Cakir, K. He, X. Xia, B. Kulis, and S. Sclaroff, “Deep metric learning to rank,” in CVPR, 2019, pp. 1861–1870

    2019

  20. [28]

    Robust and decomposable average precision for image retrieval,

    E. Ramzi, N. Thome, C. Rambour, N. Audebert, and X. Bitot, “Robust and decomposable average precision for image retrieval,” NeurIPS, vol. 34, pp. 23 569–23 581, 2021

    2021

  21. [29]

    Exploring the algorithm- dependent generalization of auprc optimization with list stability,

    P. Wen, Q. Xu, Z. Yang, Y . He, and Q. Huang, “Exploring the algorithm- dependent generalization of auprc optimization with list stability,” in NeurIPS, S. Koyejo, S. Mohamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh, Eds., vol. 35. Curran Associates, Inc., 2022, pp. 28 335– 28 349

    2022

  22. [30]

    Classification is a strong baseline for deep metric learning,

    A. Zhai and H.-Y . Wu, “Classification is a strong baseline for deep metric learning,” BMVC, 2018

    2018

  23. [31]

    No fuss distance metric learning using proxies,

    Y . Movshovitz-Attias, A. Toshev, T. K. Leung, S. Ioffe, and S. Singh, “No fuss distance metric learning using proxies,” in ICCV, 2017, pp. 360–368

    2017

  24. [32]

    Softtriple loss: Deep metric learning without triplet sampling,

    Q. Qian, L. Shang, B. Sun, J. Hu, H. Li, and R. Jin, “Softtriple loss: Deep metric learning without triplet sampling,” inICCV, 2019, pp. 6450–6458

    2019

  25. [33]

    Proxynca++: Revisiting and revitalizing proxy neighborhood component analysis,

    E. W. Teh, T. DeVries, and G. W. Taylor, “Proxynca++: Revisiting and revitalizing proxy neighborhood component analysis,” in Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XXIV 16. Springer, 2020, pp. 448–464

    2020

  26. [34]

    Unicom: Universal and compact representation learning for image retrieval,

    X. An, J. Deng, K. Yang, J. Li, Z. Feng, J. Guo, J. Yang, and T. Liu, “Unicom: Universal and compact representation learning for image retrieval,” arXiv preprint arXiv:2304.05884 , 2023

    2023 arXiv

  27. [35]

    Supervised metric learning to rank for retrieval via contextual similarity optimization,

    C. Liao, T. Tsiligkaridis, and B. Kulis, “Supervised metric learning to rank for retrieval via contextual similarity optimization,” in ICML, ser. Proceedings of Machine Learning Research, A. Krause, E. Brunskill, K. Cho, B. Engelhardt, S. Sabato, and J. Scarlett, Eds., vol. 20...

    2023

  28. [36]

    Metricformer: A unified perspective of correlation exploring in similarity learning,

    J. Yan, E. Yang, C. Deng, and H. Huang, “Metricformer: A unified perspective of correlation exploring in similarity learning,” NeurIPS, vol. 35, pp. 33 414–33 427, 2022

    2022

  29. [37]

    Causality-invariant interactive mining for cross-modal similarity learning,

    J. Yan, C. Deng, H. Huang, and W. Liu, “Causality-invariant interactive mining for cross-modal similarity learning,” IEEE TPAMI, 2024

    2024

  30. [38]

    Learning from noisy examples,

    D. Angluin and P. Laird, “Learning from noisy examples,” Machine Learning, vol. 2, pp. 343–370, 1988

    1988

  31. [39]

    Iterative learning with open-set noisy labels,

    Y . Wang, W. Liu, X. Ma, J. Bailey, H. Zha, L. Song, and S.-T. Xia, “Iterative learning with open-set noisy labels,” inCVPR, 2018, pp. 8688– 8696

    2018

  32. [40]

    Which is better for learning with noisy labels: the semi-supervised method or modeling label noise?

    Y . Yao, M. Gong, Y . Du, J. Yu, B. Han, K. Zhang, and T. Liu, “Which is better for learning with noisy labels: the semi-supervised method or modeling label noise?” in ICML. PMLR, 2023, pp. 39 660–39 673

    2023

  33. [41]

    Psnea: Pseudo- siamese network for entity alignment between multi-modal knowledge graphs,

    W. Ni, Q. Xu, Y . Jiang, Z. Cao, X. Cao, and Q. Huang, “Psnea: Pseudo- siamese network for entity alignment between multi-modal knowledge graphs,” in ACMMM, 2023, pp. 3489–3497

    2023

  34. [42]

    Positive-unlabeled learning with label distribution alignment,

    Y . Jiang, Q. Xu, Y . Zhao, Z. Yang, P. Wen, X. Cao, and Q. Huang, “Positive-unlabeled learning with label distribution alignment,” IEEE TPAMI, 2023

    2023

  35. [43]

    How does disagreement help generalization against label corruption?

    X. Yu, B. Han, J. Yao, G. Niu, I. Tsang, and M. Sugiyama, “How does disagreement help generalization against label corruption?” in ICML. PMLR, 2019, pp. 7164–7173

    2019

  36. [44]

    Improving label noise robustness with data augmentation and semi-supervised learning (student abstract),

    K. Nishi, Y . Ding, A. Rich, and T. H ¨ollerer, “Improving label noise robustness with data augmentation and semi-supervised learning (student abstract),” in AAAI, vol. 35, no. 18, 2021, pp. 15 855–15 856

    2021

  37. [45]

    Regularized contrastive partial multi-view outlier detection,

    Y . Wang, Q. Xu, Y . Jiang, S. Dai, and Q. Huang, “Regularized contrastive partial multi-view outlier detection,” in ACMMM, 2024, pp. 8711–8720

    2024

  38. [46]

    Uni- con: Combating label noise through uniform selection and contrastive learning,

    N. Karim, M. N. Rizve, N. Rahnavard, A. Mian, and M. Shah, “Uni- con: Combating label noise through uniform selection and contrastive learning,” in CVPR, 2022, pp. 9676–9686

    2022

  39. [47]

    Label-retrieval-augmented diffusion models for learning from noisy labels,

    J. Chen, R. Zhang, T. Yu, R. Sharma, Z. Xu, T. Sun, and C. Chen, “Label-retrieval-augmented diffusion models for learning from noisy labels,” NeurIPS, vol. 36, 2024

    2024

  40. [48]

    Understanding and improving early stopping for learning with noisy labels,

    Y . Bai, E. Yang, B. Han, Y . Yang, J. Li, Y . Mao, G. Niu, and T. Liu, “Understanding and improving early stopping for learning with noisy labels,” in NeurIPS, 2021

    2021

  41. [49]

    Early stopping against label noise without validation data,

    S. Yuan, L. Feng, and T. Liu, “Early stopping against label noise without validation data,” ICLR, 2024

    2024

  42. [50]

    Dm2c: Deep mixed- modal clustering,

    Y . Jiang, Q. Xu, Z. Yang, X. Cao, and Q. Huang, “Dm2c: Deep mixed- modal clustering,” NeurIPS, vol. 32, 2019

    2019

  43. [51]

    When to learn what: Deep cognitive subspace clustering,

    Y . Jiang, Z. Yang, Q. Xu, X. Cao, and Q. Huang, “When to learn what: Deep cognitive subspace clustering,” in ACMMM, 2018, pp. 718–726

    2018

  44. [52]

    Duet robust deep subspace clustering,

    Y . Jiang, Q. Xu, Z. Yang, X. Cao, and Q. Huang, “Duet robust deep subspace clustering,” in ACMMM, 2019, pp. 1596–1604

    2019

  45. [53]

    Robust distance metric learning via bayesian inference,

    D. Wang and X. Tan, “Robust distance metric learning via bayesian inference,” IEEE TIP, vol. 27, no. 3, pp. 1542–1553, 2017

    2017

  46. [54]

    Deep metric learning by online soft mining and class-aware attention,

    X. Wang, Y . Hua, E. Kodirov, G. Hu, and N. M. Robertson, “Deep metric learning by online soft mining and class-aware attention,” in AAAI, vol. 33, no. 01, 2019, pp. 5361–5368

    2019

  47. [55]

    Large-scale landmark retrieval/recognition under a noisy and diverse dataset,

    K. Ozaki and S. Yokoo, “Large-scale landmark retrieval/recognition under a noisy and diverse dataset,” arXiv preprint arXiv:1906.04087 , 2019

    1906 arXiv

  48. [56]

    Hyperbolic vision transformers: Combining improvements in metric learning,

    A. Ermolov, L. Mirvakhabova, V . Khrulkov, N. Sebe, and I. Oseledets, “Hyperbolic vision transformers: Combining improvements in metric learning,” in CVPR, 2022, pp. 7409–7419

    2022

  49. [57]

    Adaptive hierarchical similarity metric learning with noisy labels,

    J. Yan, L. Luo, C. Deng, and H. Huang, “Adaptive hierarchical similarity metric learning with noisy labels,” IEEE TIP , vol. 32, pp. 1245–1256, 2023

    2023

  50. [58]

    Unsupervised hyperbolic metric learning,

    ——, “Unsupervised hyperbolic metric learning,” in CVPR, 2021, pp. 12 465–12 474. IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 33, 2024 15

    2021

  51. [59]

    One for more: Selecting generalizable samples for generalizable reid model,

    E. Zhang, X. Jiang, H. Cheng, A. Wu, F. Yu, K. Li, X. Guo, F. Zheng, W. Zheng, and X. Sun, “One for more: Selecting generalizable samples for generalizable reid model,” in AAAI, vol. 35, no. 4, 2021, pp. 3324– 3332

    2021

  52. [60]

    Collaborative refining for person re-identification with label noise,

    M. Ye, H. Li, B. Du, J. Shen, L. Shao, and S. C. Hoi, “Collaborative refining for person re-identification with label noise,” IEEE Transactions on Image Processing , vol. 31, pp. 379–391, 2021

    2021

  53. [61]

    Noise is also useful: Negative correlation-steered latent contrastive learning,

    J. Yan, L. Luo, C. Xu, C. Deng, and H. Huang, “Noise is also useful: Negative correlation-steered latent contrastive learning,” in CVPR, 2022, pp. 31–40

    2022

  54. [62]

    Learning to purifi- cation for unsupervised person re-identification,

    L. Lan, X. Teng, J. Zhang, X. Zhang, and D. Tao, “Learning to purifi- cation for unsupervised person re-identification,” IEEE Transactions on Image Processing, 2023

    2023

  55. [63]

    Momentum contrast for unsupervised visual representation learning,

    K. He, H. Fan, Y . Wu, S. Xie, and R. Girshick, “Momentum contrast for unsupervised visual representation learning,” in CVPR, June 2020

    2020

  56. [64]

    Online deep clustering for unsupervised representation learning,

    X. Zhan, J. Xie, Z. Liu, Y .-S. Ong, and C. C. Loy, “Online deep clustering for unsupervised representation learning,” in CVPR, June 2020

    2020

  57. [65]

    Percolation and cluster distribution. i. cluster multiple labeling technique and critical concentration algorithm,

    J. Hoshen and R. Kopelman, “Percolation and cluster distribution. i. cluster multiple labeling technique and critical concentration algorithm,” Physical Review B , vol. 14, no. 8, p. 3438, 1976

    1976

  58. [66]

    Hierarchical clustering schemes,

    S. C. Johnson, “Hierarchical clustering schemes,” Psychometrika, vol. 32, no. 3, pp. 241–254, 1967

    1967

  59. [67]

    Mean shift: A robust approach toward feature space analysis,

    D. Comaniciu and P. Meer, “Mean shift: A robust approach toward feature space analysis,” IEEE TPAMI, vol. 24, no. 5, pp. 603–619, 2002

    2002

  60. [68]

    3d object representations for fine-grained categorization,

    J. Krause, M. Stark, J. Deng, and L. Fei-Fei, “3d object representations for fine-grained categorization,” in ICCV workshops, 2013, pp. 554–561

    2013

  61. [69]

    The caltech-ucsd birds-200-2011 dataset,

    C. Wah, S. Branson, P. Welinder, P. Perona, and S. Belongie, “The caltech-ucsd birds-200-2011 dataset,” 2011

    2011

  62. [70]

    Deep metric learning via lifted structured feature embedding,

    H. Oh Song, Y . Xiang, S. Jegelka, and S. Savarese, “Deep metric learning via lifted structured feature embedding,” in CVPR, 2016, pp. 4004–4012

    2016

  63. [71]

    Cleannet: Transfer learning for scalable image classifier training with label noise,

    K.-H. Lee, X. He, L. Zhang, and L. Yang, “Cleannet: Transfer learning for scalable image classifier training with label noise,” in CVPR, 2018, pp. 5447–5456

    2018

  64. [72]

    Food-101–mining discriminative components with random forests,

    L. Bossard, M. Guillaumin, and L. Van Gool, “Food-101–mining discriminative components with random forests,” in Computer Vision– ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part VI 13 . Springer, 2014, pp. 446–461

    2014

  65. [73]

    Ms-celeb-1m: A dataset and benchmark for large-scale face recognition,

    Y . Guo, L. Zhang, Y . Hu, X. He, and J. Gao, “Ms-celeb-1m: A dataset and benchmark for large-scale face recognition,” in Computer Vision– ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11-14, 2016, Proceedings, Part III 14 . Springer, 2016, pp. 87–102

    2016

  66. [74]

    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 CVPR, 2015, pp. 2691–2699

    2015

  67. [75]

    Learning with symmetric label noise: The importance of being unhinged,

    B. Van Rooyen, A. Menon, and R. C. Williamson, “Learning with symmetric label noise: The importance of being unhinged,” NeurIPS, vol. 28, 2015

    2015

  68. [76]

    Supervised contrastive learn- ing,

    P. Khosla, P. Teterwak, C. Wang, A. Sarna, Y . Tian, P. Isola, A. Maschinot, C. Liu, and D. Krishnan, “Supervised contrastive learn- ing,” NeurIPS, vol. 33, pp. 18 661–18 673, 2020

    2020

  69. [77]

    Batch normalization: Accelerating deep network training by reducing internal covariate shift,

    S. Ioffe and C. Szegedy, “Batch normalization: Accelerating deep network training by reducing internal covariate shift,” in ICML. pmlr, 2015, pp. 448–456

    2015

  70. [78]

    A metric learning reality check,

    K. Musgrave, S. Belongie, and S.-N. Lim, “A metric learning reality check,” in Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XXV 16. Springer, 2020, pp. 681–699

    2020

  71. [79]

    Paddlepaddle: An open-source deep learning platform from industrial practice,

    Y . Ma, D. Yu, T. Wu, and H. Wang, “Paddlepaddle: An open-source deep learning platform from industrial practice,” Frontiers of Data and Domputing, vol. 1, no. 1, pp. 105–115, 2019

    2019

  72. [80]

    Dynamic class queue for large scale face recognition in the wild,

    B. Li, T. Xi, G. Zhang, H. Feng, J. Han, J. Liu, E. Ding, and W. Liu, “Dynamic class queue for large scale face recognition in the wild,” in CVPR, 2021, pp. 3763–3772

    2021

  73. [81]

    Cosface: Large margin cosine loss for deep face recognition,

    H. Wang, Y . Wang, Z. Zhou, X. Ji, D. Gong, J. Zhou, Z. Li, and W. Liu, “Cosface: Large margin cosine loss for deep face recognition,” in CVPR, 2018, pp. 5265–5274

    2018

  74. [82]

    Noise-tolerant paradigm for training face recognition cnns,

    W. Hu, Y . Huang, F. Zhang, and R. Li, “Noise-tolerant paradigm for training face recognition cnns,” in CVPR, 2019, pp. 11 887–11 896

    2019

  75. [83]

    Unequal-training for deep face recognition with long-tailed noisy data,

    Y . Zhong, W. Deng, M. Wang, J. Hu, J. Peng, X. Tao, and Y . Huang, “Unequal-training for deep face recognition with long-tailed noisy data,” in CVPR, 2019, pp. 7812–7821

    2019

  76. [84]

    Co-mining: Deep face recognition with noisy labels,

    X. Wang, S. Wang, J. Wang, H. Shi, and T. Mei, “Co-mining: Deep face recognition with noisy labels,” in ICCV, 2019, pp. 9358–9367

    2019

  77. [85]

    Sub-center arcface: Boosting face recognition by large-scale noisy web faces,

    J. Deng, J. Guo, T. Liu, M. Gong, and S. Zafeiriou, “Sub-center arcface: Boosting face recognition by large-scale noisy web faces,” in Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XI 16 . Springer, 2020, pp. 741–757

    2020

  78. [86]

    Switchable k-class hyperplanes for noise-robust representation learning,

    B. Liu, G. Song, M. Zhang, H. You, and Y . Liu, “Switchable k-class hyperplanes for noise-robust representation learning,” in ICCV, 2021, pp. 3019–3028

    2021

  79. [87]

    An efficient training approach for very large scale face recognition,

    K. Wang, S. Wang, P. Zhang, Z. Zhou, Z. Zhu, X. Wang, X. Peng, B. Sun, H. Li, and Y . You, “An efficient training approach for very large scale face recognition,” in CVPR, 2022, pp. 4083–4092. Zhipeng Yu received the B.E. degree in communica- tion engineering and the M.E degre...

    2022

  80. [2023]

    Her research interests include machine learning and computer vision

    She is currently a postdoctoral research fellow with University of Chinese Academy of Sciences. Her research interests include machine learning and computer vision. She has authored or coauthored several academic papers in international journals and conferences including T-PAM...

    1999

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.