Pith. sign in

REVIEW 4 major objections 6 minor 52 references

GUESS: Generative Uncertainty Ensemble for Self Supervision

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper claims that replacing the identity target in whitening-based self-supervised learning with a data-derived target from autoencoder correlations improves representations and sets new baselines on several benchmarks.

desk verdict GUESS has a plausible pseudo-whitening idea, but the loss in Eq. (1) differs from the pseudocode's, so the reported 'new baseline' is not tied to a well-defined method. read the letter →

arxiv 2412.02896 v1 pith:J72FLHWH submitted 2024-12-03 cs.LG cs.CV

Salman Mohamadi , Gianfranco Doretto , Donald A. Adjeroh This is my paper · ORCID
classification cs.LGcs.CV
keywords self-supervisedlearningpseudo-whiteningredundancyreductionuncertaintyrepresentationensemblegenerative-discriminativelossdataaugmentationinvariance
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

Self-supervised learning usually trains a network to map two augmented views of the same image to similar representations, and recent whitening methods go further by forcing the cross-correlation of the two views' embeddings to be the identity matrix. This paper argues that such strict, blind invariance enforcement discards useful information and can hurt downstream tasks. It proposes GUESS, which instead learns a pseudo-whitening target: the diagonal of the target is still 1, but the off-diagonal entries come from the cross-correlation of two autoencoders fed the same augmented views, making the loss data-dependent and preserving view-specific structure. An ensemble of such blocks, each trained on its own distorted views, outperforms or stays competitive with Barlow-Twins, Whitening-MSE, SimCLR, BYOL, SimSiam, and SwAV across four classification benchmarks, with the largest gains at ensemble sizes three and five. If the claim holds, uncertainty-aware invariance enforcement is a practical and general alternative to aggressive whitening in self-supervised learning.

What carries the argument

The generative uncertainty block is the unit that carries the argument. It couples a twin encoder-projector pair with a twin autoencoder pair: the encoders produce the representation to be learned, while the autoencoders, trained for reconstruction on the same augmented views, produce a data-dependent uncertainty signal. The cross-correlation of the autoencoder latents, with its diagonal set to 1, replaces the identity matrix as the target in the whitening loss, so the loss no longer forces off-diagonal correlations to zero and instead lets the data decide what is redundant. The second machinery piece is the ensemble: $M$ such blocks, each with its own augmentation stream and its own reconstruction-trained autoencoders, whose classifiers vote at test time. The efficient variant reuses one branch and an auto-correlation target to halve the compute, turning a fixed decorrelation constraint into a per-sample, per-distortion constraint.

What would settle it

Train GUESS-1 on ImageNet with the autoencoder-derived target replaced by a fixed random matrix with the same off-diagonal distribution, keeping all other settings identical; if accuracy stays near GUESS-1's 75.6, the data-dependence of the target is not the cause of the gain, while a drop toward Barlow-Twins' 73.6 would confirm the autoencoder signal is the active ingredient.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the optimal whitening target in redundancy-reduction self-supervised learning is not the identity matrix but a data-dependent matrix shaped by the input distortions. For each block, GUESS computes two cross-correlation matrices: one from the two encoder-projector branches and one from the latent spaces of two autoencoders. The loss $L_w = \sum_i (1-C_{ii})^2 + \beta \sum_{i \neq j} (C_{ij} - C_{1,ij})^2$ pulls the encoder cross-correlation toward a target whose off-diagonal entries come from the autoencoders rather than from zero, so the optimizer is uncertain about which correlations are redundant. The architecture wraps this block in an ensemble of $M$ such blocks, each fed its own augmented views, and combines their classifiers by majority vote at test time; an efficient variant replaces cross-correlation with auto-correlation, halving the compute. The paper reports that even the one-block version outperforms all baselines on ImageNet linear evaluation with 75.6 top-1 accuracy, and that three- and five-block ensembles push accuracy further on the linear-evaluation benchmarks.

Load-bearing premise

The load-bearing premise is that the cross-correlation of two autoencoders' latent vectors is a meaningful and stable measure of augmentation-induced uncertainty; if that matrix mostly reflects reconstruction artifacts or training noise, the pseudo-whitening target carries no useful signal and the reported gains would come from the ensemble or from regularization alone.

Editorial extensions

If this is right

  • If GUESS is correct, the identity matrix should not be treated as the universal whitening target; a data-derived pseudo-whitening target is a direct drop-in replacement for the decorrelation term in whitening-based SSL losses.
  • The ensemble scaling results imply that adding independently augmented blocks generally improves linear-evaluation accuracy, at least up to the tested size of five.
  • The efficient ensemble results imply that the improvement does not require the doubled compute of the full architecture: GUESS-1-E stays within about 0.1 to 0.2 accuracy points of GUESS-1 while adding roughly no computational overhead over Barlow-Twins.
  • The heavy-augmentation ablation implies that the data-dependent target is more forgiving than deterministic invariance: under the heavy augmentation protocol GUESS-3 keeps 50.11 top-1 accuracy on Tiny ImageNet, where SimSiam falls to 44.11.
  • The ablation without autoencoder pre-training implies that the uncertainty target must be learned before it is used: skipping the 250-epoch pre-training drops Tiny ImageNet accuracy from 50.91 to 48.16 and slows convergence.

Reading between the lines

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

  • A direct extension the paper leaves implicit: the same autoencoder-derived target could be plugged into other redundancy-reduction or contrastive losses, and testing that transfer would reveal whether the mechanism is tied to the Barlow-Twins-style loss or is a general property of data-dependent whitening.
  • The Appendix C claim that the surrogate objective "maintains the same global optimum" as the information-bottleneck objective is asserted without proof; a formal derivation or a counterexample would settle whether the gains come from a principled objective change or from optimization-side effects.
  • If the autoencoder target is truly measuring augmentation-induced uncertainty, then its quality should degrade smoothly with autoencoder capacity and pre-training duration, which would make the framework a candidate for online or adaptive uncertainty estimates rather than fixed pre-trained targets.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes GUESS, a self-supervised learning framework that replaces the strict whitening target of Barlow-Twins with a data-dependent "pseudo-whitening" target derived from the cross-correlation of a pair of autoencoder latents, and that combines this loss with an ensemble of blocks, each trained on its own augmented views. The authors report linear-evaluation and transfer-learning results on CIFAR10/100, Tiny ImageNet, ImageNet, VOC0712, and COCO, and claim that GUESS, even with one block, sets a new baseline on several benchmarks. An efficient variant replaces the cross-correlation with an auto-correlation to halve computational cost.

Significance. If the empirical claims hold, GUESS would be a meaningful contribution to non-contrastive SSL: it introduces a concrete mechanism for data-dependent invariance enforcement, evaluates it across six datasets and three downstream tasks, and provides an efficient variant with comparable cost to Barlow-Twins. The paper also ships a substantial ablation study, including sensitivity to beta, robustness to heavy augmentation, and the effect of autoencoder pre-training. However, the central 'new baseline' claim is currently undermined by an inconsistency between the stated loss and the provided pseudocode, by overstatements in the text relative to the reported tables, and by the absence of error bars or multiple seeds. These issues must be resolved before the empirical claims can be accepted.

major comments (4)
  1. [Section III-B, Eq. (1) and Algorithm 1] The loss defined in Eq. (1) and the loss minimized in Algorithm 1 are not the same objective. In Eq. (1), the network cross-correlation matrix C is pulled toward the autoencoder cross-correlation C1, with the off-diagonal mismatch weighted by beta. In Algorithm 1, C1 denotes the network correlation, C2 denotes the autoencoder correlation, and the loss is (C1 - (eye(D) + beta*C2))^2, which expands to sum_i (1 - C1_ii)^2 + sum_{i != j} (C1_ij - beta*C2_ij)^2. These differ in which matrix is the target, in the placement of beta, and in the treatment of diagonal entries. Consequently, the reported 'new baseline' results cannot be attributed to the loss stated in Eq. (1), and the method is not reproducible as written. Please align the pseudocode with Eq. (1), or vice versa, and state explicitly which objective was actually optimized in the experiments.
  2. [Section IV-D.2 and Table II] The text claims that 'GUESS-5 outperforms the state-of-the-art' on CIFAR10 when pre-training on Tiny ImageNet, but Table II lists BYOL at 93.17 and GUESS-5 at 92.93 for exactly this setting. Similarly, Section IV-D.1 states that GUESS-1 and GUESS-1-E 'consistently improved over all baselines' on CIFAR100 under linear evaluation, yet Table I shows W-MSE4 at 67.64 and GUESS-1-E at 67.53. These overstatements contradict the paper's own tables and should be corrected, with the narrative aligned to the reported numbers.
  3. [Section IV-C and Tables I-III] No error bars, standard deviations, or multiple seeds are reported for any experiment. Many of the claimed improvements are small (for example, CIFAR100 GUESS-1 67.67 vs W-MSE4 67.64, and Tiny ImageNet GUESS-5 52.33 vs SimSiam 51.66), so single-run numbers are insufficient to support the 'new baseline' claims. Please report mean and standard deviation over at least three independent runs, and state how many seeds were used for the reproduced baselines.
  4. [Appendix C, after Eq. (15)] The assertion that the surrogate objective 'maintains the same global optimum' as Eq. (15) is made without proof. This is a nontrivial statement relating a Frobenius-norm surrogate to a log-determinant information-bottleneck objective, and it is especially doubtful because the surrogate uses a non-identity off-diagonal target while the IB objective involves covariance determinants. Either provide a derivation or explicitly label this as a conjecture.
minor comments (6)
  1. [Section IV-B] The motivating experiment on pseudo-whitening reports only a 0.47% gain (50.51% vs 50.04%) over Barlow-Twins, and the Gaussian-noise variant gives 50.11%; given these small margins, the language describing this as supporting 'pseudo-whitening' should be tempered or supplemented with additional evidence.
  2. [Algorithm 1 and Eq. (1)] The symbol C is used both for the pseudo-whitening matrix in Algorithm 1 and for the network cross-correlation matrix in Eq. (1); this notation conflict makes the inconsistency between the two formulations harder to detect and should be fixed.
  3. [Table III] The abbreviation 'NP' in 'GUESS-1 (NP)' is not defined in the table or text; it should be expanded (e.g., 'no autoencoder pre-training').
  4. [Section IV-D.2] There is a typo 'GUERSS-1' instead of 'GUESS-1'.
  5. [Appendix A and Eq. (2)] Eq. (2) and Appendix A's Eq. (6) are identical; one of the two should be removed to avoid duplication.
  6. [References and Appendix B] Reference [25] is a podcast rather than a peer-reviewed source; please cite the underlying literature instead. Also, 'Lecunn' should be 'LeCun' in Appendix B.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: GUESS's 'new baseline' claim is supported by external benchmark comparisons, not by a derivation that reduces to its own inputs.

full rationale

The central claim of the paper is empirical: GUESS is reported to outperform six baselines on linear, transfer, detection, and segmentation benchmarks. Those outcomes are not generated from the loss or architecture by construction; they are measured against external datasets under standard protocols. The pseudo-whitening target C = eye(D) + beta*C2 is computed from the autoencoder latents on the same augmented batch, which makes the training objective self-referential in the usual SSL sense, but it does not force any downstream accuracy number. The self-citations in the manuscript are contextual (deep active learning ensemble sampling, FUSSL, joint-mutual-information work, and the author's dissertation) and are not used to justify the 'new baseline' claim; no uniqueness theorem or fitted parameter from prior author work is invoked. Two passages deserve correctness flags, not circularity flags: Section III-B Eq. (1) and Algorithm 1 define different objectives (Eq. (1) uses the projector-head correlation C and the autoencoder correlation C1 in a swapped and differently weighted role from the pseudocode's (C1 - (eye + beta*C2))^2), and Appendix C asserts without proof that the Frobenius-norm surrogate 'maintains the same global optimum' as the information-bottleneck objective. Both are reproducibility/support problems; neither makes a reported result equivalent to its input by definition.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central claim depends on several hand-chosen hyperparameters and an ad hoc assumption that autoencoder cross-correlation provides a meaningful uncertainty target. The information bottleneck connection is asserted without proof. No new physical or conceptual entities are introduced.

free parameters (5)
  • alpha = 0.2
    Reported as best value in Section IV.A, but Algorithm 1 does not use alpha in the loss, so its role is unclear.
  • beta = 0.01
    Reported as best value in Section IV.A; controls the target off-diagonal magnitude beta*C2. Ablation in Figure 3 shows insensitivity over small variations.
  • augmentation probability p = random distribution per view
    Controls the allocation of augmented views to ensemble blocks in Section IV.A; chosen according to prior SSL protocols.
  • ensemble size M = 1, 3, 5
    Selected by the authors to demonstrate scaling; not a derived quantity.
  • autoencoder pre-training epochs = 250
    Set in Section IV.A; ablation shows pre-training matters, without pre-training accuracy drops significantly (50.91 to 48.16 for beta=0.01).
assumptions (4)
  • standard math Z-score normalization of embeddings and cross-correlation matrices are valid for the SSL objective.
    Used throughout Algorithm 1 and Section III.B to define C1 and C2.
  • domain assumption Invariance to data augmentation is a desirable learning signal for representation learning.
    The paper's motivation rests on this assumption, inherited from Barlow-Twins and prior SSL methods, stated in the introduction.
  • ad hoc to paper Autoencoder cross-correlation is a good proxy for data-derived uncertainty in invariance enforcement.
    Introduced in Section III.A without external justification; the only support is the small initial experiment in Section IV.B.
  • ad hoc to paper The surrogate objective in Eq. (1) / Algorithm 1 maintains the same global optimum as the information bottleneck objective in Eq. (15).
    Asserted in Appendix C with no proof; Appendix C also contains an unfinished sentence about SwAV.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GUESS: Generative Uncertainty Ensemble for Self Supervision." pith.science (2026). https://pith.science/paper/J72FLHWH

@misc{pith2026241202896,
  author       = {Pith},
  title        = {Pith review of: GUESS: Generative Uncertainty Ensemble for Self Supervision},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/J72FLHWH}},
  note         = {Machine review of arXiv:2412.02896}
}
read the original abstract

Self-supervised learning (SSL) frameworks consist of pretext task, and loss function aiming to learn useful general features from unlabeled data. The basic idea of most SSL baselines revolves around enforcing the invariance to a variety of data augmentations via the loss function. However, one main issue is that, inattentive or deterministic enforcement of the invariance to any kind of data augmentation is generally not only inefficient, but also potentially detrimental to performance on the downstream tasks. In this work, we investigate the issue from the viewpoint of uncertainty in invariance representation. Uncertainty representation is fairly under-explored in the design of SSL architectures as well as loss functions. We incorporate uncertainty representation in both loss function as well as architecture design aiming for more data-dependent invariance enforcement. The former is represented in the form of data-derived uncertainty in SSL loss function resulting in a generative-discriminative loss function. The latter is achieved by feeding slightly different distorted versions of samples to the ensemble aiming for learning better and more robust representation. Specifically, building upon the recent methods that use hard and soft whitening (a.k.a redundancy reduction), we introduce a new approach GUESS, a pseudo-whitening framework, composed of controlled uncertainty injection, a new architecture, and a new loss function. We include detailed results and ablation analysis establishing GUESS as a new baseline.

Figures

Figures reproduced from arXiv: 2412.02896 by the authors.

Figure 1
Figure 1. Schematic depiction of the proposed GUESS framework, with random allocation of augmented views to the blocks of ensemble- [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. More efficient GUESS framework. For more efficiency, the building [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Sensitivity to beta, for CIFAR10, as shown the top-1 linear accuracy [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 29 canonical work pages

  1. [1]

    Self-supervised visual feature learning with deep neural networks: A survey,

    L. Jing and Y . Tian, “Self-supervised visual feature learning with deep neural networks: A survey,” IEEE transactions on pattern analysis and machine intelligence, vol. 43, no. 11, pp. 4037–4058, 2020

    2020

  2. [2]

    A unified visual information preservation framework for self-supervised pre-training in medical image analysis,

    H.-Y . Zhou, C. Lu, C. Chen, S. Yang, and Y . Yu, “A unified visual information preservation framework for self-supervised pre-training in medical image analysis,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023

    work page 2023

  3. [3]

    Deep active ensemble sam- pling for image classification,

    S. Mohamadi, G. Doretto, and D. Adjeroh, “Deep active ensemble sam- pling for image classification,” in Proceedings of the Asian Conference on Computer Vision , 2022, pp. 4531–4547

    work page 2022

  4. [4]

    Deep bayesian active learning, a brief survey on recent advances,

    S. Mohamadi and H. Amindavar, “Deep bayesian active learning, a brief survey on recent advances,” arXiv preprint arXiv:2012.08044 , 2020

    arXiv 2012

  5. [5]

    A survey on deep semi-supervised learning,

    X. Yang, Z. Song, I. King, and Z. Xu, “A survey on deep semi-supervised learning,” IEEE Transactions on Knowledge and Data Engineering , vol. 35, no. 9, pp. 8934–8954, 2022

    2022

  6. [6]

    Active uncertainty representation learning: Toward more label efficiency in deep learning,

    S. Mohamadi, “Active uncertainty representation learning: Toward more label efficiency in deep learning,” 2024

    work page 2024

  7. [7]

    Barlow twins: Self-supervised learning via redundancy reduction,

    J. Zbontar, L. Jing, I. Misra, Y . LeCun, and S. Deny, “Barlow twins: Self-supervised learning via redundancy reduction,” in International Conference on Machine Learning . PMLR, 2021, pp. 12 310–12 320

    2021

  8. [8]

    Fussl: Fuzzy uncertain self supervised learning,

    S. Mohamadi, G. Doretto, and D. A. Adjeroh, “Fussl: Fuzzy uncertain self supervised learning,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , 2023, pp. 2799–2808

    work page 2023

Show all 52 references
  1. [9]

    More synergy, less redundancy: Exploiting joint mutual in- formation for self-supervised learning,

    ——, “More synergy, less redundancy: Exploiting joint mutual in- formation for self-supervised learning,” in 2023 IEEE International Conference on Image Processing (ICIP) . IEEE, 2023, pp. 1390–1394

    2023

  2. [10]

    Learning where to learn in cross-view self-supervised learning,

    L. Huang, S. You, M. Zheng, F. Wang, C. Qian, and T. Yamasaki, “Learning where to learn in cross-view self-supervised learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 14 451–14 460

    2022

  3. [11]

    Representation learning with contrastive predictive coding,

    A. v. d. Oord, Y . Li, and O. Vinyals, “Representation learning with contrastive predictive coding,” arXiv preprint arXiv:1807.03748 , 2018. 9

    2018 arXiv

  4. [12]

    Contrastive multiview coding,

    Y . Tian, D. Krishnan, and P. Isola, “Contrastive multiview coding,” in European conference on computer vision . Springer, 2020, pp. 776–794

    2020

  5. [13]

    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 Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 9729–9738

    2020

  6. [14]

    A simple framework for contrastive learning of visual representations,

    T. Chen, S. Kornblith, M. Norouzi, and G. Hinton, “A simple framework for contrastive learning of visual representations,” in International conference on machine learning . PMLR, 2020, pp. 1597–1607

    2020

  7. [15]

    Learning representations by maximizing mutual information across views,

    P. Bachman, R. D. Hjelm, and W. Buchwalter, “Learning representations by maximizing mutual information across views,” Advances in neural information processing systems , vol. 32, 2019

    2019

  8. [16]

    Bootstrap your own latent-a new approach to self-supervised learning,

    J.-B. Grill, F. Strub, F. Altch ´e, C. Tallec, P. Richemond, E. Buchatskaya, C. Doersch, B. Avila Pires, Z. Guo, M. Gheshlaghi Azar et al. , “Bootstrap your own latent-a new approach to self-supervised learning,” Advances in neural information processing systems , vol. 33, pp....

    2020

  9. [17]

    Exploring simple siamese representation learning,

    X. Chen and K. He, “Exploring simple siamese representation learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 15 750–15 758

    2021

  10. [18]

    Understanding self-supervised learn- ing dynamics without contrastive pairs,

    Y . Tian, X. Chen, and S. Ganguli, “Understanding self-supervised learn- ing dynamics without contrastive pairs,” in International Conference on Machine Learning . PMLR, 2021, pp. 10 268–10 278

    2021

  11. [19]

    Whitening for self-supervised representation learning,

    A. Ermolov, A. Siarohin, E. Sangineto, and N. Sebe, “Whitening for self-supervised representation learning,” in International Conference on Machine Learning . PMLR, 2021, pp. 3015–3024

    2021

  12. [20]

    Dense contrastive learning for self-supervised visual pre-training,

    X. Wang, R. Zhang, C. Shen, T. Kong, and L. Li, “Dense contrastive learning for self-supervised visual pre-training,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 3024–3033

    2021

  13. [21]

    Propagate yourself: Exploring pixel-level consistency for unsupervised visual representation learning,

    Z. Xie, Y . Lin, Z. Zhang, Y . Cao, S. Lin, and H. Hu, “Propagate yourself: Exploring pixel-level consistency for unsupervised visual representation learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 16 684–16 693

    2021

  14. [22]

    Spatially consistent representation learning,

    B. Roh, W. Shin, I. Kim, and S. Kim, “Spatially consistent representation learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 1144–1153

    2021

  15. [23]

    Region similarity representation learning,

    T. Xiao, C. J. Reed, X. Wang, K. Keutzer, and T. Darrell, “Region similarity representation learning,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2021, pp. 10 539–10 548

    2021

  16. [24]

    Self-supervised models are continual learners,

    E. Fini, V . G. T. da Costa, X. Alameda-Pineda, E. Ricci, K. Alahari, and J. Mairal, “Self-supervised models are continual learners,” in Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 9621–9630

    2022

  17. [25]

    Lex fridman podcast, (MIT AI podcast). episode # 258

    “Lex fridman podcast, (MIT AI podcast). episode # 258.”

  18. [26]

    Using self- supervised learning can improve model robustness and uncertainty,

    D. Hendrycks, M. Mazeika, S. Kadavath, and D. Song, “Using self- supervised learning can improve model robustness and uncertainty,” Advances in neural information processing systems , vol. 32, 2019

    2019

  19. [27]

    On the uncertainty of self-supervised monocular depth estimation,

    M. Poggi, F. Aleotti, F. Tosi, and S. Mattoccia, “On the uncertainty of self-supervised monocular depth estimation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2020, pp. 3227–3237

    2020

  20. [28]

    Exploiting unlabeled data in cnns by self-supervised learning to rank,

    X. Liu, J. Van De Weijer, and A. D. Bagdanov, “Exploiting unlabeled data in cnns by self-supervised learning to rank,” IEEE transactions on pattern analysis and machine intelligence , vol. 41, no. 8, pp. 1862–1878, 2019

    2019

  21. [29]

    Self-supervised low-light image enhancement using discrepant untrained network priors,

    J. Liang, Y . Xu, Y . Quan, B. Shi, and H. Ji, “Self-supervised low-light image enhancement using discrepant untrained network priors,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 32, no. 11, pp. 7332–7345, 2022

    2022

  22. [30]

    Self-supervised learning by estimating twin class distribution,

    F. Wang, T. Kong, R. Zhang, H. Liu, and H. Li, “Self-supervised learning by estimating twin class distribution,” IEEE Transactions on Image Processing, 2023

    2023

  23. [31]

    Adversarial feature learn- ing,

    J. Donahue, P. Kr ¨ahenb¨uhl, and T. Darrell, “Adversarial feature learn- ing,” arXiv preprint arXiv:1605.09782 , 2016

    2016 arXiv

  24. [32]

    Unsupervised learning of visual features by contrasting cluster assign- ments,

    M. Caron, I. Misra, J. Mairal, P. Goyal, P. Bojanowski, and A. Joulin, “Unsupervised learning of visual features by contrasting cluster assign- ments,” Advances in Neural Information Processing Systems , vol. 33, pp. 9912–9924, 2020

    2020

  25. [33]

    Deep clustering for unsupervised learning of visual features,

    M. Caron, P. Bojanowski, A. Joulin, and M. Douze, “Deep clustering for unsupervised learning of visual features,” in Proceedings of the European conference on computer vision (ECCV) , 2018, pp. 132–149

    2018

  26. [34]

    Emerging properties in self-supervised vision transformers,

    M. Caron, H. Touvron, I. Misra, H. J ´egou, J. Mairal, P. Bojanowski, and A. Joulin, “Emerging properties in self-supervised vision transformers,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 9650–9660

    2021

  27. [35]

    Improved deep metric learning with multi-class n-pair loss objective,

    K. Sohn, “Improved deep metric learning with multi-class n-pair loss objective,” Advances in neural information processing systems , vol. 29, 2016

    2016

  28. [36]

    Unsupervised learning of visual representations using videos,

    X. Wang and A. Gupta, “Unsupervised learning of visual representations using videos,” in Proceedings of the IEEE international conference on computer vision , 2015, pp. 2794–2802

    2015

  29. [37]

    Learning multiple layers of features from tiny images,

    A. Krizhevsky, G. Hinton et al. , “Learning multiple layers of features from tiny images,” 2009

    2009

  30. [38]

    Tiny imagenet visual recognition challenge,

    Y . Le and X. Yang, “Tiny imagenet visual recognition challenge,” CS 231N, vol. 7, no. 7, p. 3, 2015

    2015

  31. [39]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2016, pp. 770–778

    2016

  32. [40]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980 , 2014

    2014 arXiv

  33. [41]

    solo-learn: A library of self-supervised methods for visual representation learning

    V . G. T. da Costa, E. Fini, M. Nabi, N. Sebe, and E. Ricci, “solo-learn: A library of self-supervised methods for visual representation learning.” J. Mach. Learn. Res. , vol. 23, pp. 56–1, 2022

    2022

  34. [42]

    Directional self-supervised learning for heavy image augmentations,

    Y . Bai, Y . Yang, W. Zhang, and T. Mei, “Directional self-supervised learning for heavy image augmentations,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 16 692–16 701

    2022

  35. [43]

    Shuffle and learn: unsupervised learning using temporal order verification,

    I. Misra, C. L. Zitnick, and M. Hebert, “Shuffle and learn: unsupervised learning using temporal order verification,” in European conference on computer vision . Springer, 2016, pp. 527–544

    2016

  36. [44]

    Contrastive and non-contrastive self- supervised learning recover global and local spectral embedding meth- ods,

    R. Balestriero and Y . LeCun, “Contrastive and non-contrastive self- supervised learning recover global and local spectral embedding meth- ods,” arXiv preprint arXiv:2205.11508 , 2022

    2022 arXiv

  37. [45]

    Uncertainty-aware self-supervised learning of spatial perception tasks,

    M. Nava, A. Paolillo, J. Guzzi, L. M. Gambardella, and A. Giusti, “Uncertainty-aware self-supervised learning of spatial perception tasks,” IEEE Robotics and Automation Letters , vol. 6, no. 4, pp. 6693–6700, 2021

    2021

  38. [46]

    Credal self-supervised learning,

    J. Lienen and E. H ¨ullermeier, “Credal self-supervised learning,” Ad- vances in Neural Information Processing Systems , vol. 34, pp. 14 370– 14 382, 2021

    2021

  39. [47]

    Cliquecnn: Deep unsupervised exemplar learning,

    M. A. Bautista, A. Sanakoyeu, E. Tikhoncheva, and B. Ommer, “Cliquecnn: Deep unsupervised exemplar learning,” Advances in Neural Information Processing Systems , vol. 29, 2016

    2016

  40. [48]

    Invariant information clustering for unsupervised image classification and segmentation,

    X. Ji, J. F. Henriques, and A. Vedaldi, “Invariant information clustering for unsupervised image classification and segmentation,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2019, pp. 9865–9874

    2019

  41. [49]

    The information bottleneck method,

    N. Tishby, F. C. Pereira, and W. Bialek, “The information bottleneck method,” arXiv preprint physics/0004057 , 2000. 10 APPENDIX A EFFICIENT ENSEMBLE WITH AUTO -CORRELATION As discussed in the paper, we propose efficient ensemble , in which we reduce the computational complexi...

    2000 arXiv

  42. [50]

    and later SimSiam [17] devised a type of loss which is needless of negative instances while avoiding representation collapse. In particular BYOL shows that negative instances are dispensable and the framework surprisingly avoids the representation collapse caused by utilizing ...

  43. [51]

    , suggests that non-contrastive loss functions are generally more preferable due to better error bound on downstream tasks. Aside from above mentioned methods which directly in- volve features, a different set of approaches based on clustering [32], [33], [33], [34] primarily ...

  44. [52]

    this work proposes the use of credal sets to model uncertainty in pseudo-labels and hence reduce calibration errors in SSL approach

    translates the concept of credal sets (set of probability distributions) to SSL in order provide model uncertainty in pseudo-labels in low regime labeled data. this work proposes the use of credal sets to model uncertainty in pseudo-labels and hence reduce calibration errors i...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.