Pith. sign in

REVIEW 5 major objections 6 minor 39 references

Direct Coloring for Self-Supervised Enhanced Feature Decoupling

T0 review · 5 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 Adding a 'direct coloring' loss that regresses an intermediate layer's cross-correlation matrix toward a VAE-computed target improves self-supervised learning, reaching 76.1% top-1 linear accuracy on ImageNet with ResNet50.

desk verdict Plausible coloring regularizer with a circular MAP story and a missing null-target control; worth a careful review but not as-is. read the letter →

arxiv 2412.02109 v1 pith:BAP6HVNM submitted 2024-12-03 cs.CV

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

This paper claims that self-supervised representation learning improves when the network is also trained to 'color' its features: an intermediate layer's cross-correlation matrix is driven, by a squared-error loss, toward a target matrix computed from the latent spaces of two variational autoencoders trained on augmented views. That coloring term is added to a Barlow-Twins-style whitening loss, giving a total loss $\mathcal{L} = \mathcal{L}_W + \lambda \mathcal{L}_C$ with a small $\lambda$. On ImageNet with ResNet50, the authors report 76.1% top-1 linear accuracy after 1000 epochs, 0.8 points above the best prior baseline (DINO at 75.3%), with faster convergence at 100 and 400 epochs; they also report gains on CIFAR10/100, VOC detection, and COCO segmentation. The motivation is that augmentation already decouples sparse useful features from dense less-useful ones, and direct coloring is meant to push that decoupling further while the whitening stage handles dimensional collapse. The authors further argue, from a constrained-optimization view of the combined loss, that the coloring term lowers the chance of complete collapse.

What carries the argument

The load-bearing object is the desired cross-correlation matrix $E$, computed from the latent spaces of a pair of variational autoencoders trained on the two augmented views, with entries $E_{ij} = \sum_n z^{(1)}_{n,i} z^{(2)}_{n,j} / \sqrt{\sum_n (z^{(1)}_{n,i})^2} \sqrt{\sum_n (z^{(2)}_{n,j})^2}$. The coloring loss $\mathcal{L}_C = \sum_{i,j} (C_{ij} - E_{ij})^2$ regresses the cross-correlation $C$ of intermediate 'coloring head' outputs toward $E$, promoting the decoupled feature structure encoded in $E$ at an early layer; the whitening loss $\mathcal{L}_W$ then decorrelates the final embedding. Direct coloring skips the classical whitening-then-coloring transformation and drives $C$ straight to $E$ by gradient descent, which the authors argue is computationally faster and does not assume a multivariate Gaussian. The combined loss is presented as a MAP estimate with a Gaussian prior $p(\Theta) = \mathcal{N}(\Theta | E, \sigma^2)$, so $E$ serves as both the 'Bayesian prior' and the coloring target.

What would settle it

Re-run the ImageNet-100 pipeline with the target matrix $E$ replaced by (a) the identity matrix, (b) a column-shuffled version of the VAE matrix, or (c) autoencoder latents instead of VAE latents. If top-1 accuracy barely changes, the specific content of $E$ is not what drives the gain; if accuracy drops sharply, the VAE-derived structure matters. A second check is to repeat the 1000-epoch ImageNet comparison over several seeds to see whether the 0.8-point lead over DINO is stable.

Watch

Extended reading notes

Core claim

In the paper's own terms, the central discovery is that direct coloring of the cross-correlation matrix enhances feature decoupling in self-supervised learning and improves downstream representations. Two symmetric networks produce two views of each sample; a 'coloring head' attached to an intermediate layer (layer 16 of ResNet18, layer 46 of ResNet50) computes a cross-correlation matrix $C$, and the coloring loss $\mathcal{L}_C = \sum_i \sum_j (C_{ij} - E_{ij})^2$ pulls it toward a desired matrix $E$ whose entries are the cross-correlations between the latent vectors of two VAEs trained on the same augmented views, as in Eq. 5. The final projector heads are whitened in the Barlow-Twins style, $\mathcal{L}_W = \sum_i (1 - W_{ii})^2 + \alpha \sum_{i \neq j} W_{ij}^2$. Because coloring is applied directly to $C$ rather than through the classical whiten-then-color pipeline, the authors call it direct coloring and argue it is faster and makes no Gaussian assumption. They report 76.1% top-1 linear accuracy on ImageNet with ResNet50 at 1000 epochs, exceeding DINO by 0.8 points, and that the coloring head alone accounts for a 1.24-point gain on ImageNet-100 (80.93% vs. 79.69% without it).

Load-bearing premise

The load-bearing premise is that the cross-correlation matrix computed from two variational autoencoders trained on the same augmented views is the right target for feature decoupling; the paper offers no independent reason why this particular data-derived matrix, rather than some other correlation structure, is the desired one.

Editorial extensions

If this is right

  • Any SSL pipeline that already whitens or decorrelates its embedding can absorb the coloring term with a small weight $\lambda$ and improve linear-evaluation accuracy; the paper reports +0.8 points over DINO on ImageNet top-1 at 1000 epochs.
  • Training converges faster in epoch count: the paper reports 69.6% top-1 on ImageNet at 100 epochs and 73.2% at 400 epochs, both above the previous baselines at the same budget.
  • Coloring transfers to non-whitening methods: added to SimSiam with $\lambda = 0.01$ it raises ImageNet-100 top-1 from 77.17% to 78.40%, while $\lambda = 0.05$ degrades it to 72.5%.
  • A simplified auto-correlation variant on a single network halves the computation with a small accuracy cost (80.64% vs. 80.93% on ImageNet-100).
  • Coloring reduces the chance of complete collapse: the variance of the whitening head's normalized output on ImageNet-100 is 0.97 with coloring versus 0.68 without.

Reading between the lines

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

  • Editorial: the target matrix $E$ is derived from VAEs trained on the same augmented data, so the method is effectively distilling the data's own correlation structure back into the network; the gains may depend on the VAE latents capturing the sparse-versus-dense feature split the paper assumes.
  • Editorial: the MAP argument in Section 3.3 is self-referential, because both the prior $p(\Theta) = \mathcal{N}(\Theta | E, \sigma^2)$ and the coloring likelihood $\prod_{i,j} \mathcal{N}(C_{ij} | E_{ij}, \sigma^2)$ peak at the same matrix $E$; the Bayesian framing does not independently justify why $E$ is the right target.
  • Editorial: a direct test of the mechanism would replace $E$ with hand-designed targets (identity, block-diagonal, or column-shuffled matrices) and measure accuracy; if the specific structure of $E$ does not matter, the benefit is just an extra regression constraint.
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 a 'direct coloring' loss for self-supervised learning (SSL). The method trains two symmetric networks on augmented views; an intermediate layer of each encoder feeds a coloring projector head, and the loss matches the cross-correlation matrix C of the coloring head outputs to a target matrix E obtained from the latent spaces of two variational autoencoders (VAEs) trained on the same augmented data. This coloring term is added to a Barlow-Twins-style whitening loss (L = LW + λLC, Eq. (1)). The paper reports top-1 linear accuracy of 76.1% on ImageNet with ResNet50 after 1000 epochs, surpassing the listed baselines, and claims faster convergence plus reduced complete collapse. It also presents a maximum-a-posteriori (MAP) argument in Sec. 3.3 and ablations on head location, projector dimension, target matrix choice, and a simplified auto-correlation variant.

Significance. The method is simple, modular, and broadly evaluated: results are reported on ImageNet, CIFAR10/100, Tiny ImageNet, VOC0712, and COCO, against contrastive, non-contrastive, clustering, whitening, and transformer-based baselines. The ablations in Secs. 5.1-5.4 and the supplementary, especially the comparison of VAE vs AE targets and the SimSiam upgrade experiment, are useful in mapping the design space. However, the paper's core theoretical justification is invalid as written, and the empirical evidence does not yet show that the specific content of the target matrix E is what drives the improvements. The strengths are the breadth of evaluation and the modularity of the proposal; the main weaknesses are the circular MAP derivation, the missing null-target control, the underspecified VAE training protocol, and the absence of statistical error bars.

major comments (5)
  1. [Section 3.3, Eqs. (6)-(11)] The MAP derivation is not a valid justification of the loss. The prior p(Θ)=N(Θ|E,σ²) is defined over the model parameters Θ but is centered at E, the target cross-correlation matrix, which is not a parameter of the network; the coloring likelihood in Eq. (8) is also centered at E for each Cij. The two factors are thus the same quadratic penalty written twice, so the 'Bayesian prior' carries no independent information and the derivation simply restates the coloring loss. Furthermore, Eq. (1) contains no term of the form ||Θ-E||², so the negative log-prior in Eq. (11) does not correspond to any term in L. I recommend removing this justification or replacing it with a derivation in which the prior and likelihood are genuinely distinct.
  2. [Section 5.3] The ablation of the desired matrix E only compares a VAE-derived target to an autoencoder-derived target (80.93% vs 80.21% on ImageNet-100). This does not control for the possibility that any fixed target matrix, or simply the presence of the auxiliary coloring head and extra loss term, yields a similar gain. Without ablations against a trivial target such as the identity matrix, a fixed random matrix, or a zero target, the paper's central claim that the VAE-derived E encodes the desired feature decoupling is not established. Please add these controls, or demonstrate by another means that the per-entry structure of E matters.
  3. [Section 4.2.1 and supplementary] The VAE training protocol is severely underspecified. The text states only that 'the architecture of the VAE is based on ResNet18 or ResNet50' and refers to the supplementary for details, but the supplementary does not provide the VAE loss function, number of training epochs, optimizer, hyperparameters, or the exact augmentation regime used to train the VAEs. Because the target E is entirely produced by these VAEs, the method cannot be reproduced or its sensitivity to VAE training assessed. This information must be supplied.
  4. [Tables 1-4] All results are reported as single runs without error bars, standard deviations, or significance tests. The claimed improvements are small in several cases (0.8% on ImageNet at 1000 epochs, 1.02% on CIFAR10), so without run-to-run variance estimates the improvements cannot be distinguished from noise. Please report mean ± std over at least three seeds for the main comparisons, and state the number of seeds.
  5. [Section 3.2] The paper asserts that the VAE-derived cross-correlation matrix E encodes the desired 'feature decoupling', but provides no direct evidence or formal argument for this link. The downstream linear accuracy does not by itself show that the representations exhibit the claimed sparse/dense feature structure, nor that E's specific entries are what cause the effect. A comparison of E with trivial matrices (identity, random) and an analysis of the learned cross-correlation structure would strengthen this central premise.
minor comments (6)
  1. [Throughout] There are numerous typos and spacing issues (e.g., 'functionτ' in Sec. 2, 'learing' in Sec. 3.2, 'CIAFAR100' in the Table 2 caption, 'AP100' in the Table 3 header); the manuscript needs a careful proofread.
  2. [Section 5.4 and figures] Figure numbering is inconsistent: Sec. 5.4 refers to 'Fig. 3' for the auto-correlation architecture, but the figure labeled 'Figure 3' in the text is the sensitivity-to-λ plot at the start of the supplementary. Please renumber all figures.
  3. [Section 4.2.3] The phrase 'the value of λ in the loss function (12)' should refer to Eq. (1) (or Eqs. (1)-(2)), since Eq. (12) in the supplementary belongs to the auto-correlation variant.
  4. [Section 3.3] The phrase 'aligns with the terms in Equation 12 and 13' should refer to the main loss equations; the numbering is confusing because the main text does not have an Eq. (12)/(13).
  5. [Section 5.5 and supplementary] The complete-collapse experiment reports the variance of the normalized whitening-head output (0.97 with coloring vs 0.68 without), but the variance definition and normalization are not described; please specify the formula and the batch used.
  6. [Abstract and Section 4.4] The claim of 'faster convergence' is based on the epoch-100/400/1000 columns of Table 1, which is not a direct convergence measure; a training-curve plot would be more appropriate.

Circularity Check

1 steps flagged · score 5.0 of 10

MAP 'Bayesian prior' is the coloring likelihood restated: both Gaussians are centered at the same VAE-derived E, so the derivation reduces to the loss it claims to justify.

  1. self definitional [Section 3.3 (MAP analysis), Eqs. (6)-(11), especially Eqs. (7)-(8)]
    "here we have a prior in the form of the target colored cross-correlation matrix E from V AEs, which specifies the desired correlation structure between the augmented views. Hence the prior is Gaussian distribution with mean E and a certain variance σ2 as follows: p(Θ) = N (Θ|E, σ2) ... pcolor(X|Θ) = ∏_i ∏_j N (Cij |Eij, σ2)"

    In the MAP objective (Eq. 6), the prior p(Θ) and the coloring likelihood p_color(X|Θ) are both Gaussian densities centered at the same matrix E: the prior is N(Θ|E, σ²) and the coloring likelihood is N(C_ij|E_ij, σ²). The posterior therefore contains two identical squared-error factors in (C_ij - E_ij), so taking -log gives the coloring loss LC of Eq. (4) twice (plus the unrelated whitening terms). No independent prior information is injected: the 'Bayesian prior' is just a second copy of the coloring likelihood. The text never defines a generative model connecting Θ to C or W, so the MAP 'solution' is a notational restatement of the loss, not a derivation of it.

full rationale

The one clear circular step is the MAP analysis in Section 3.3: the claimed Bayesian prior has the same mean as the coloring likelihood, so the derivation is self-definitional and reduces to the loss it pretends to justify. This is a genuine circularity in the paper's theoretical framing. However, the primary empirical claim—that adding the coloring loss improves SSL accuracy—is not circular in the statistical sense: the target E is precomputed from auxiliary VAEs, and the reported ImageNet/CIFAR/VOC/COCO numbers are external benchmark evaluations against standard baselines, not predictions forced by a fitted parameter. The ablation in Section 5.3 compares VAE-derived vs AE-derived targets and omits a null-target control (identity, random, or self-derived target); that is a real experimental gap in the causal attribution to E, but it is a missing control rather than a circular step. The paper's self-citations ([23]-[26]) are not load-bearing for the coloring mechanism. Because the circularity is confined to a derivation that restates the loss, while the empirical results retain independent content, the score is 5 rather than 8 or 10.

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

The method does not introduce new physical entities or mediators. The free parameters are standard SSL hyperparameters plus a VAE-derived target matrix E, which is essentially a fitted quantity. The axioms include the augmentation/feature-decoupling assumption from prior literature and several assumptions specific to this paper, including the validity of E and the correctness of the MAP derivation.

free parameters (5)
  • coloring loss weight λ = 0.05 (static, ImageNet); tuned on ImageNet-100 over range 0.005-0.08
    Weight balancing the coloring loss LC against the whitening loss LW. Set to 0.05 for main experiments after ablation on ImageNet-100; large values lead to dimensional collapse (Supplementary A.1).
  • whitening off-diagonal weight α = 0.01
    Trades off diagonal vs off-diagonal terms in the whitening loss LW. Fixed at 0.01 following the general Barlow Twins-style setting, but the exact value is a hyperparameter.
  • coloring head location (layer) = 16 for ResNet18, 46 for ResNet50
    The layer from which features are fed to the coloring heads. Chosen via ablation on ImageNet-100; layer 10 and 17 both degrade accuracy, indicating sensitivity.
  • projector output dimension = 2048 for ImageNet, 1024 for CIFAR10/100 and Tiny ImageNet
    The size of the whitening/coloring head outputs and VAE latent space. Ablation on ImageNet-100 shows accuracy varies from 74.31% (512) to 81.66% (4096), so the choice is performance-driven.
  • target cross-correlation matrix E = computed from VAE latents on the training set
    The desired coloring target is estimated by training two VAEs on augmented views and computing their latent cross-correlation. This is a data-derived matrix that the coloring loss fits to; it is the central quantity the method depends on.
assumptions (4)
  • domain assumption Proper augmentation decouples sparse (useful) features from dense (less useful) features, and enhancing this decoupling improves SSL.
    Invoked in the Introduction and Section 3.1, citing Wen and Li [35,36]. The whole motivation for coloring rests on this assumed mechanism of augmentation.
  • ad hoc to paper The cross-correlation between VAE latents of augmented views encodes the desired feature decoupling and is a valid target E.
    Section 3.2 states E is computed from VAEs and used as the desired cross-correlation, but no theoretical result or external evidence shows that this particular matrix is optimal for downstream tasks. This is the load-bearing premise of the method.
  • ad hoc to paper The MAP prior p(Θ)=N(Θ|E,σ2) and the coloring likelihood p_color=N(C|E,σ2) are independent factors, allowing the MAP objective to be written as their product.
    Section 3.3, Eqs. (7)-(11). The prior and likelihood are both centered at E, so they are the same term. This double-counting is an internal inconsistency in the derivation.
  • domain assumption The coloring constraint g has non-zero gradient away from constant representations, so adding it as a constraint prevents complete collapse.
    Section 5.5 and Appendix E argue that complete collapse is 'certainly not a solution to g', but this is asserted without formal proof. The empirical variance measurement supports it, but the theoretical claim is a sketch.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Direct Coloring for Self-Supervised Enhanced Feature Decoupling." pith.science (2026). https://pith.science/paper/BAP6HVNM

@misc{pith2026241202109,
  author       = {Pith},
  title        = {Pith review of: Direct Coloring for Self-Supervised Enhanced Feature Decoupling},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BAP6HVNM}},
  note         = {Machine review of arXiv:2412.02109}
}
read the original abstract

The success of self-supervised learning (SSL) has been the focus of multiple recent theoretical and empirical studies, including the role of data augmentation (in feature decoupling) as well as complete and dimensional representation collapse. While complete collapse is well-studied and addressed, dimensional collapse has only gain attention and addressed in recent years mostly using variants of redundancy reduction (aka whitening) techniques. In this paper, we further explore a complementary approach to whitening via feature decoupling for improved representation learning while avoiding representation collapse. In particular, we perform feature decoupling by early promotion of useful features via careful feature coloring. The coloring technique is developed based on a Bayesian prior of the augmented data, which is inherently encoded for feature decoupling. We show that our proposed framework is complementary to the state-of-the-art techniques, while outperforming both contrastive and recent non-contrastive methods. We also study the different effects of coloring approach to formulate it as a general complementary technique along with other baselines.

Figures

Figures reproduced from arXiv: 2412.02109 by the authors.

Figure 1
Figure 1. Left:Schematic diagram of the proposed framework. For a given sample, two augmented views are generated and fed to the [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Simpler architecture with auto-correlation instead of [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Sensitivity to λ. A. λ A.1. Sensitivity to λ A range of experiments on ImageNet-100 with different values for λ is presented here. Note that the value for α is set to 10−2 for all experiments and the pre-training was performed only for 250 epochs. The top-1 accuracy for dif￾ferent values of λ is depicted in [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 25 canonical work pages

  1. [1]

    Mine your own view: Self-supervised learning through across-sample prediction

    Mehdi Azabou, Mohammad Gheshlaghi Azar, Ran Liu, Chi- Heng Lin, Erik C Johnson, Kiran Bhaskaran-Nair, Max Dabagia, Bernardo Avila-Pires, Lindsey Kitchell, Keith B Hengen, et al. Mine your own view: Self-supervised learning through across-sample prediction. arXiv preprint arXiv:2102.10106, 2021. 2

    arXiv 2021

  2. [2]

    Direc- tional self-supervised learning for heavy image augmenta- tions

    Yalong Bai, Yifan Yang, Wei Zhang, and Tao Mei. Direc- tional self-supervised learning for heavy image augmenta- tions. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , pages 16692–16701,

  3. [3]

    Deep clustering for unsupervised learning of visual features

    Mathilde Caron, Piotr Bojanowski, Armand Joulin, and Matthijs Douze. Deep clustering for unsupervised learning of visual features. In Proceedings of the European confer- ence on computer vision (ECCV) , pages 132–149, 2018. 2

    work page 2018

  4. [4]

    Unsupervised learning of visual features by contrasting cluster assignments

    Mathilde Caron, Ishan Misra, Julien Mairal, Priya Goyal, Pi- otr Bojanowski, and Armand Joulin. Unsupervised learning of visual features by contrasting cluster assignments. Ad- vances in Neural Information Processing Systems , 33:9912– 9924, 2020. 2, 5

    work page 2020

  5. [5]

    Emerg- ing properties in self-supervised vision transformers

    Mathilde Caron, Hugo Touvron, Ishan Misra, Herv ´e J´egou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. Emerg- ing properties in self-supervised vision transformers. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 9650–9660, 2021. 5

    work page 2021

  6. [6]

    A simple framework for contrastive learning of visual representations

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Ge- offrey Hinton. A simple framework for contrastive learning of visual representations. In International conference on ma- chine learning, pages 1597–1607. PMLR, 2020. 2, 3, 5, 6

    work page 2020

  7. [7]

    Exploring simple siamese rep- resentation learning

    Xinlei Chen and Kaiming He. Exploring simple siamese rep- resentation learning. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition , pages 15750–15758, 2021. 2, 5

    work page 2021

  8. [8]

    Understanding generalized whitening and col- oring transform for universal style transfer

    Tai-Yin Chiu. Understanding generalized whitening and col- oring transform for universal style transfer. In Proceedings of the IEEE/CVF International Conference on Computer Vi- sion, pages 4452–4460, 2019. 3

    work page 2019

Show all 39 references
  1. [9]

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

    Victor Guilherme Turrisi da Costa, Enrico Fini, Moin Nabi, Nicu Sebe, and Elisa Ricci. solo-learn: A library of self- supervised methods for visual representation learning. J. Mach. Learn. Res., 23:56–1, 2022. 5, 6

    2022

  2. [10]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition, pages 248–255. Ieee, 2009. 5

    2009

  3. [11]

    Whitening for self-supervised representation learning

    Aleksandr Ermolov, Aliaksandr Siarohin, Enver Sangineto, and Nicu Sebe. Whitening for self-supervised representation learning. In International Conference on Machine Learning, pages 3015–3024. PMLR, 2021. 2, 3, 4, 5, 6

    2021

  4. [12]

    The pascal visual object classes (voc) challenge

    Mark Everingham, Luc Van Gool, Christopher KI Williams, John Winn, and Andrew Zisserman. The pascal visual object classes (voc) challenge. International journal of computer vision, 88:303–308, 2009. 5

    2009

  5. [13]

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

    Jean-Bastien Grill, Florian Strub, Florent Altch ´e, Corentin Tallec, Pierre Richemond, Elena Buchatskaya, Carl Doersch, Bernardo Avila Pires, Zhaohan Guo, Mohammad Ghesh- laghi Azar, et al. Bootstrap your own latent-a new approach to self-supervised learning. Advances in neur...

    2020

  6. [14]

    Momentum contrast for unsupervised visual rep- resentation learning

    Kaiming He, Haoqi Fan, Yuxin Wu, Saining Xie, and Ross Girshick. Momentum contrast for unsupervised visual rep- resentation learning. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 9729–9738, 2020. 7

    2020

  7. [15]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceed- ings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016. 5

    2016

  8. [16]

    Whitening and coloring transformations for multivariate gaussian data

    Maliha Hossain. Whitening and coloring transformations for multivariate gaussian data. A slecture partly based on the ECE662 Spring, 2014. 4

    2014

  9. [17]

    On feature decorrelation in self- supervised learning

    Tianyu Hua, Wenxiao Wang, Zihui Xue, Sucheng Ren, Yue Wang, and Hang Zhao. On feature decorrelation in self- supervised learning. In Proceedings of the IEEE/CVF Inter- national Conference on Computer Vision, pages 9598–9608,

  10. [18]

    Self-supervised visual fea- ture learning with deep neural networks: A survey

    Longlong Jing and Yingli Tian. Self-supervised visual fea- ture learning with deep neural networks: A survey. IEEE transactions on pattern analysis and machine intelligence , 43(11):4037–4058, 2020. 1

    2020

  11. [19]

    Adam: A method for stochastic optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 ,

    arXiv

  12. [20]

    Learning multiple layers of features from tiny images

    Alex Krizhevsky, Geoffrey Hinton, et al. Learning multiple layers of features from tiny images. 2009. 5

    2009

  13. [21]

    Tiny imagenet visual recognition challenge

    Ya Le and Xuan Yang. Tiny imagenet visual recognition challenge. CS 231N, 7(7):3, 2015. 5

    2015

  14. [22]

    Microsoft coco: Common objects in context

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll´ar, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In Computer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceeding...

    2014

  15. [23]

    Active uncertainty representation learn- ing: Toward more label efficiency in deep learning

    Salman Mohamadi. Active uncertainty representation learn- ing: Toward more label efficiency in deep learning. 2024. 1

    2024

  16. [24]

    Deep active ensemble sampling for image classification

    Salman Mohamadi, Gianfranco Doretto, and Don Adjeroh. Deep active ensemble sampling for image classification. In Proceedings of the Asian Conference on Computer Vision , pages 4531–4547, 2022. 1

    2022

  17. [25]

    Fussl: Fuzzy uncertain self supervised learning

    Salman Mohamadi, Gianfranco Doretto, and Donald A Ad- jeroh. Fussl: Fuzzy uncertain self supervised learning. In Proceedings of the IEEE/CVF Winter Conference on Appli- cations of Computer Vision, pages 2799–2808, 2023. 1

    2023

  18. [26]

    More synergy, less redundancy: Exploiting joint mu- tual information for self-supervised learning

    Salman Mohamadi, Gianfranco Doretto, and Donald A Ad- jeroh. More synergy, less redundancy: Exploiting joint mu- tual information for self-supervised learning. In 2023 IEEE International Conference on Image Processing (ICIP), pages 1390–1394. IEEE, 2023. 1

    2023

  19. [27]

    Enhancement of sinusoids in colored noise and the whitening performance of exact least squares predictors

    Arye Nehorai and Martin Morf. Enhancement of sinusoids in colored noise and the whitening performance of exact least squares predictors. IEEE Transactions on Acoustics, Speech, and Signal Processing, 30(3):353–363, 1982. 2 9

    1982

  20. [28]

    Unsu- pervised domain adaptation using full-feature whitening and colouring

    Subhankar Roy, Aliaksandr Siarohin, and Nicu Sebe. Unsu- pervised domain adaptation using full-feature whitening and colouring. In Image Analysis and Processing–ICIAP 2019: 20th International Conference, Trento, Italy, September 9– 13, 2019, Proceedings, Part II 20 , pages 225...

    2019

  21. [29]

    Backdoor attacks on self- supervised learning

    Aniruddha Saha, Ajinkya Tejankar, Soroush Abbasi Kooh- payegani, and Hamed Pirsiavash. Backdoor attacks on self- supervised learning. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition , pages 13337–13346, 2022. 2

    2022

  22. [30]

    Whitening and coloring batch transform for gans

    Aliaksandr Siarohin, Enver Sangineto, and Nicu Sebe. Whitening and coloring batch transform for gans. arXiv preprint arXiv:1806.00420, 2018. 2, 3

    2018 arXiv

  23. [31]

    Con- trastive multiview coding

    Yonglong Tian, Dilip Krishnan, and Phillip Isola. Con- trastive multiview coding. In European conference on com- puter vision, pages 776–794. Springer, 2020. 1

    2020

  24. [32]

    What makes for good views for contrastive learning? Advances in Neural Infor- mation Processing Systems, 33:6827–6839, 2020

    Yonglong Tian, Chen Sun, Ben Poole, Dilip Krishnan, Cordelia Schmid, and Phillip Isola. What makes for good views for contrastive learning? Advances in Neural Infor- mation Processing Systems, 33:6827–6839, 2020. 1, 2

    2020

  25. [33]

    The information bottleneck method

    Naftali Tishby, Fernando C Pereira, and William Bialek. The information bottleneck method. arXiv preprint physics/0004057, 2000. 1

    2000 arXiv

  26. [34]

    Un- supervised representation learning by invariance propaga- tion

    Feng Wang, Huaping Liu, Di Guo, and Sun Fuchun. Un- supervised representation learning by invariance propaga- tion. Advances in Neural Information Processing Systems , 33:3510–3520, 2020. 2

    2020

  27. [35]

    Toward understanding the fea- ture learning process of self-supervised contrastive learning

    Zixin Wen and Yuanzhi Li. Toward understanding the fea- ture learning process of self-supervised contrastive learning. In International Conference on Machine Learning , pages 11112–11122. PMLR, 2021. 1, 2, 3

    2021

  28. [36]

    The mechanism of predic- tion head in non-contrastive self-supervised learning

    Zixin Wen and Yuanzhi Li. The mechanism of predic- tion head in non-contrastive self-supervised learning. arXiv preprint arXiv:2205.06226, 2022. 1, 2, 3

    2022 arXiv

  29. [37]

    Seed the views: Hi- erarchical semantic alignment for contrastive representation learning

    Haohang Xu, Xiaopeng Zhang, Hao Li, Lingxi Xie, Wen- rui Dai, Hongkai Xiong, and Qi Tian. Seed the views: Hi- erarchical semantic alignment for contrastive representation learning. IEEE Transactions on Pattern Analysis and Ma- chine Intelligence, 2022. 2

    2022

  30. [38]

    Photorealistic style transfer via wavelet transforms

    Jaejun Yoo, Youngjung Uh, Sanghyuk Chun, Byeongkyu Kang, and Jung-Woo Ha. Photorealistic style transfer via wavelet transforms. In Proceedings of the IEEE/CVF Inter- national Conference on Computer Vision, pages 9036–9045,

  31. [39]

    Barlow twins: Self-supervised learning via redundancy reduction

    Jure Zbontar, Li Jing, Ishan Misra, Yann LeCun, and St´ephane Deny. Barlow twins: Self-supervised learning via redundancy reduction. In International Conference on Ma- chine Learning, pages 12310–12320. PMLR, 2021. 2, 3, 4, 5, 6, 8 10 Figure 3. Sensitivity to λ. A. λ A.1. Sens...

    2021

Pith tools

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