Pith. sign in

REVIEW 5 major objections 6 minor 55 references

Rethinking the Sample Relations for Few-Shot Classification

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

Pith's one-line read Modeling sample relations at three granularities — exact consistency within a sample, closeness within a class, separation between classes — yields state-of-the-art few-shot classification accuracy and a pre-trained backbone that improves…

desk verdict A sensible decomposition of sample relations with a useful pretraining recipe, but the SOTA claim is not statistically secured; the method is competitive and deserves revision, not desk rejection. read the letter →

arxiv 2501.13418 v1 pith:RJJECFOO submitted 2025-01-23 cs.CV cs.AI

classification cs.CVcs.AI
keywords few-shotclassificationcontrastivelearningsamplerelationstransformationconsistencyclasspre-trainingfeature
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper claims that existing few-shot contrastive pre-training fails because it models all sample relations with one recipe, ignoring that a transformed view of an image is semantically identical to it, while a same-class image is only similar and a different-class image is dissimilar. It proposes Multi-Grained Relation Contrastive Learning (MGRCL), which splits relations into intra-sample, intra-class, and inter-class, enforces the first with label-distribution alignment (TCL) and the latter two with a relative contrastive loss (CCL). A sympathetic reader would care because, if correct, a simple single-stage pre-training method matches or exceeds two-stage meta-learners and can be dropped into existing few-shot methods as a better frozen backbone.

What carries the argument

The central objects are the two auxiliary losses added to a standard classification backbone. TCL operates on the classifier's softmax outputs, softened by temperature $\tau_1 = 4.0$, and penalizes the Jensen-Shannon divergence between the prediction for the original image and the prediction for each of its transformed versions, enforcing identity rather than mere similarity at the label level. CCL operates on global-pooled feature embeddings, drawing a positive same-class feature from a memory bank, treating features of the other base classes as negatives, and optimizing a temperature-$\tau_2=0.1$ softmax over cosine similarities; the memory bank is updated with momentum $r = 0.99$. A self-supervised branch that predicts which transformation was applied completes the base learner. Together the losses separate the three relation types that unsupervised and supervised contrastive learning conflate.

What would settle it

Train the same base learner with and without TCL on a dataset where one of the used augmentations (for instance, 90-degree rotation on digit images) flips the semantic label; if the TCL-trained model does not beat the no-TCL baseline on held-out 5-way 1-shot tasks, the premise that label outputs capture semantic content and that alignment preserves it is not supported.

Watch

Extended reading notes

Core claim

MGRCL claims that the three kinds of sample relations — same sample under different transformations, different samples of the same class, and samples of different classes — have different strengths of semantic similarity and therefore need different modeling. Transformation Consistency Learning (TCL) enforces exact semantic consistency by aligning temperature-smoothed predicted label distributions of an image and its transformed versions with Jensen-Shannon divergence, while Class Contrastive Learning (CCL) uses a memory bank and a softmax over cosine similarities to keep each sample closer to a same-class sample than to samples of other classes. On this scheme the paper reports 5-way 1-shot accuracies of 69.57% on miniImageNet, 72.98% on tieredImageNet, 78.54% on CIFAR-FS, and 86.14% on CUB, and shows that using the resulting network as a frozen pre-trained backbone improves FEAT, Meta-Baseline, and STV AE on the same benchmarks.

Load-bearing premise

The load-bearing premise is that the classifier's predicted label distribution faithfully represents a sample's semantic content, so forcing identical label outputs across transformations preserves semantics rather than discarding useful information.

Editorial extensions

If this is right

  • A single fully supervised pre-training phase on base classes, without episode sampling or meta-tuning, is enough to reach state-of-the-art 5-way 1-shot accuracy on all four tested benchmarks.
  • Existing two-stage methods such as FEAT, Meta-Baseline, and STV AE inherit substantial gains simply by starting from the MGRCL backbone rather than a plain supervised one.
  • On fine-grained datasets like CUB, where class differences are small, the class-pushing CCL component contributes more than the transformation-alignment TCL component.
  • Unsupervised contrastive losses that treat every other image as negative can hurt few-shot pre-training because they push same-class samples apart; MGRCL avoids that by contrasting against a same-class positive from a memory bank.
  • Accuracy improvements come without extra network layers beyond a small MLP for the self-supervised transformation-prediction branch, keeping inference identical to standard feature-extraction-based few-shot classifiers.

Reading between the lines

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

  • Editorial inference: If label-output alignment is doing the work, TCL may also act as a strong regularizer against overfitting on small base datasets, which would predict larger gains when base-class data is scarce; this is not tested in the paper.
  • Editorial inference: The same three-way relation split could be applied beyond images — to text or audio — where augmentation families differ, and the method's success would depend on whether label distributions there faithfully reflect semantic content.
  • Editorial inference: A direct test of the paper's motivation would be to replace the memory bank with class prototypes; if performance holds, the relative contrastive formulation, not the bank, is the essential ingredient.
  • Editorial inference: Because the paper reports that aggressive augmentations such as Sobel edge detection are included, an interesting stress test is whether TCL suppresses texture-based features that are useful for some novel classes but absent after edge filtering.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The paper proposes Multi-Grained Relation Contrastive Learning (MGRCL), a pre-training method for few-shot classification that models sample relations at three granularities: intra-sample relations (Transformation Consistency Learning, TCL, which aligns temperature-smoothed label distributions of a sample and its augmented versions via Jensen-Shannon divergence), and intra-class/inter-class relations (Class Contrastive Learning, CCL, a memory-bank-based contrastive loss that pulls a sample toward a homogenous sample and pushes it away from samples of all other classes). The base learner jointly optimizes a classification loss, a self-supervised transformation-prediction loss, TCL, and CCL. The method is evaluated on miniImageNet, tieredImageNet, CIFAR-FS, and CUB with a ResNet-12 backbone, reporting 5-way 1-shot accuracies of 69.57%, 72.98%, 78.54%, and 86.14%, respectively, and it is claimed to be state-of-the-art in 5-way 1-shot classification. The paper also shows that using MGRCL as a pre-trained backbone improves the two-stage methods FEAT, Meta-Baseline, and STV AE, and it provides component ablations, hyperparameter sweeps, comparisons to UnSupCon/SupCon, and t-SNE visualizations.

Significance. If the empirical claims hold, MGRCL is a simple and transferable pre-training recipe with a conceptually clean decomposition of sample relations, and the plug-and-play gains for FEAT, Meta-Baseline, and STV AE are practically useful. The ablations in Table 5 and the same-base-learner comparison in Fig. 5 give explicit credit to each designed component, and Sections 5.2-5.3 candidly acknowledge the computational overhead and memory-bank scaling cost. Balanced against these strengths, the headline state-of-the-art claim is not statistically secured on miniImageNet and tieredImageNet (overlapping confidence intervals), the hyperparameters are selected on the headline benchmark (Section 4.4), and no code or complete augmentation protocol is provided. These are fixable within the manuscript's scope, so I regard the contribution as potentially solid but in need of substantial additional evidence before it can be accepted.

major comments (5)
  1. [4.2, Tables 1-2] The claim that MGRCL achieves state-of-the-art 5-way 1-shot accuracy is not supported by the reported statistics on two of the four benchmarks. On miniImageNet, Ours (69.57±0.45) is 0.20 points above PAL (69.37±0.64), and on tieredImageNet, Ours (72.98±0.51) is 0.30 points above ESPT (72.68±0.22); the quoted 95% confidence intervals overlap in both cases. Since Section 4.1.5 states that the intervals are computed over 2,000 sampled tasks, and no variance across training seeds is reported anywhere, the pairwise margins do not demonstrate a statistically meaningful advantage. Please report means and standard deviations over multiple training seeds with a significance test, or qualify the state-of-the-art wording.
  2. [4.4, Table 6] The loss weights α and β are selected by grid search on miniImageNet (Table 6), and the temperature parameters τ1 and τ2 are tuned on miniImageNet (Fig. 4); the headline miniImageNet results in Tables 1 and 5 coincide exactly with the best grid point (α=1.0, β=0.1, 69.57). The reported miniImageNet number is thus the maximum over the searched grid, which is a selection-on-the-headline-benchmark effect. The paper should either verify the chosen hyperparameters on the other three benchmarks (e.g., by reporting the grid landscape on CUB), or report the results of a configuration fixed without reference to the test benchmark.
  3. [4.1.4 and code availability] The quantitative claims are not reproducible from the manuscript. Section 4.1.4 lists the augmentation types ('three scaling transformations, three rotation transformations, one random erasing, one graying, and one Sobel edge detection') without any parameters, the self-supervised MLP in Section 4.1.2 is underspecified (hidden dimension, activation), and no code or checkpoints for MGRCL are provided - the only URL in the paper (Section 4.5) points to the SupCon repository used for baseline comparisons. Please provide the full augmentation protocol and training details and release code, so that the central empirical claims can be reproduced and checked across training seeds.
  4. [3.4] The TCL component rests on the premise stated in Section 3.4 that 'the label output can represent the semantic content of the sample.' This premise is load-bearing: for aggressive augmentations such as Sobel edge detection or rotation, forcing exact Jensen-Shannon consistency of label distributions could suppress useful invariant or equivariant features. The aggregate gains in Table 5 are consistent with the premise but do not isolate it. Please add an ablation that removes each augmentation type from the TCL objective one at a time, or that compares TCL against a feature-level consistency loss, to verify that the semantic-consistency assumption holds for every transformation used.
  5. [4.3, Table 5] The 5-way 1-shot Baseline in Table 5 is 66.78±0.43 on miniImageNet, which exactly matches the FEAT result reported in Table 1 (66.78±0.20), even though Section 4.3 describes the baseline as 'the same as RFS' with added augmented samples, and RFS-distill is reported as 64.82±0.60 in Table 1. Please clarify the exact training recipe that produced the baseline and state explicitly whether it is an RFS-style cross-entropy model, a FEAT model, or another configuration; every incremental improvement in Table 5 is measured from this value.
minor comments (6)
  1. [5.1 and 4.6] The heading 'Limitions' in Section 5.1 should be 'Limitations', and 'effevtiveness' in Section 4.6 should be 'effectiveness'.
  2. [1, 4.1.1, Table 3] The dataset name is spelled 'CIF AR-FS' with a space in Sections 1, 4.1.1, and the Table 3 caption; use 'CIFAR-FS' consistently.
  3. [4.6] The phrase 'the second of Figure 6b' is unclear; it should be 'the second row of Figure 6b' or similar.
  4. [References] Reference [37] is missing a comma between the authors and has a spacing error in the venue name ('onInformation theory'); reference [14] lacks volume and page information.
  5. [4.2] The sentence 'outperforming the second-best results of 0.69% and 0.73%' should read 'outperforming the second-best results by 0.69 and 0.73 percentage points.'
  6. [4.5] The footnote about using the SupCon code for both the unsupervised and supervised baselines should state explicitly which loss function (SimCLR-style or SupCon-style) is applied as 'UnSupCon' and 'SupCon' in Fig. 5, since both are implemented with the same repository.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: MGRCL's reported gains are empirical measurements, not consequences of its loss definitions.

full rationale

MGRCL is presented as a training objective rather than as a formal derivation whose predictions are constructed from assumptions. The losses Lcls, Lss, Ltcl, and Lccl are defined directly in Sections 3.3-3.5, and the few-shot accuracies in Tables 1-4 are measured after training; there is no step in which a reported result equals an input or a fitted parameter by construction. The TCL component does rely on the stated assumption that label outputs represent semantic content, but the paper uses this as a design choice for the loss, not as a derivation of the empirical outcome. Self-citations to STV AE [14] and HandCrafted [20] appear only as baselines or combination partners alongside external methods such as FEAT and Meta-Baseline, so no load-bearing claim reduces to those citations. The hyperparameter grid search on miniImageNet in Section 4.4 is a statistical selection concern rather than circularity, because the reported 69.57% is an evaluated accuracy and is not forced to equal the selected values of α, β, τ1, or τ2 by the loss equations; tieredImageNet, CIFAR-FS, and CUB are evaluated with fixed hyperparameters and provide independent evidence. No enumerated circular step can be substantiated from the paper's equations or citations.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The method introduces two loss weights and two temperature parameters, all selected on miniImageNet, which is also the dataset where headline accuracy is reported. The memory bank momentum is inherited from prior work. No invented entities are postulated; the memory bank is an established technique.

free parameters (5)
  • alpha (TCL loss weight) = 1.0
    Chosen via grid search on miniImageNet (Section 4.4, Table 6).
  • beta (CCL loss weight) = 0.1
    Chosen via grid search on miniImageNet; larger values degrade accuracy (Section 4.4).
  • tau_1 (TCL temperature) = 4.0
    Tuned on miniImageNet; results are insensitive between 1 and 10 (Figure 4a).
  • tau_2 (CCL temperature) = 0.1
    Tuned on miniImageNet following SimCLR/SupCon (Figure 4b).
  • r (memory bank momentum) = 0.99
    Momentum for the memory bank update, set following prior work [18].
assumptions (3)
  • domain assumption Transformation Consistency Learning requires that different augmentations preserve exactly the same semantic content, and that such content is expressible as a full predicted label distribution.
    Section 3.4 states 'The label output can represent the semantic content of the sample.' This is a modeling premise, not proven, and aggressive augmentations could carry information beyond the class label.
  • domain assumption The memory bank updated with momentum r = 0.99 yields feature embeddings accurate enough for CCL's positive and negative sampling during training.
    Eq. (9) defines the update; the paper does not analyze staleness or its effect on the contrastive gradient.
  • domain assumption The three-way relation partition is sufficient: no additional relation types are needed, and the two losses do not conflict.
    The method's success depends on this design choice; the paper provides intuitive arguments and ablations, not a theoretical guarantee.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Rethinking the Sample Relations for Few-Shot Classification." pith.science (2026). https://pith.science/paper/RJJECFOO

@misc{pith2026250113418,
  author       = {Pith},
  title        = {Pith review of: Rethinking the Sample Relations for Few-Shot Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RJJECFOO}},
  note         = {Machine review of arXiv:2501.13418}
}
read the original abstract

Feature quality is paramount for classification performance, particularly in few-shot scenarios. Contrastive learning, a widely adopted technique for enhancing feature quality, leverages sample relations to extract intrinsic features that capture semantic information and has achieved remarkable success in Few-Shot Learning (FSL). Nevertheless, current few-shot contrastive learning approaches often overlook the semantic similarity discrepancies at different granularities when employing the same modeling approach for different sample relations, which limits the potential of few-shot contrastive learning. In this paper, we introduce a straightforward yet effective contrastive learning approach, Multi-Grained Relation Contrastive Learning (MGRCL), as a pre-training feature learning model to boost few-shot learning by meticulously modeling sample relations at different granularities. MGRCL categorizes sample relations into three types: intra-sample relation of the same sample under different transformations, intra-class relation of homogenous samples, and inter-class relation of inhomogeneous samples. In MGRCL, we design Transformation Consistency Learning (TCL) to ensure the rigorous semantic consistency of a sample under different transformations by aligning predictions of input pairs. Furthermore, to preserve discriminative information, we employ Class Contrastive Learning (CCL) to ensure that a sample is always closer to its homogenous samples than its inhomogeneous ones, as homogenous samples share similar semantic content while inhomogeneous samples have different semantic content. Our method is assessed across four popular FSL benchmarks, showing that such a simple pre-training feature learning method surpasses a majority of leading FSL methods. Moreover, our method can be incorporated into other FSL methods as the pre-trained model and help them obtain significant performance gains.

Figures

Figures reproduced from arXiv: 2501.13418 by the authors.

Figure 1
Figure 1. In this figure, shapes and colors represent different samples and classes respec [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Network architecture of our model, which contains a CNN backbone [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Effects of hyper-parameters α and β on miniImageNet. (a) α. (b) β [PITH_FULL_IMAGE:figures/full_fig_p020_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Effects of hyper-parameters τ1 and τ2 on miniImageNet. (a) τ1. (b) τ2. performance when α and β is set to 1.0 and 0.1 respectively, As shown in [PITH_FULL_IMAGE:figures/full_fig_p021_4.png]
Figure 5
Figure 5. Figure 5: Comparison to the base learner with Unsupervised Contrastive Learning or [PITH_FULL_IMAGE:figures/full_fig_p022_5.png]
Figure 6
Figure 6. Figure 6: t-SNE visualization results for 5 randomly selected novel classes of miniIma [PITH_FULL_IMAGE:figures/full_fig_p023_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

55 extracted references · 47 canonical work pages

  1. [1]

    Krizhevsky, I

    A. Krizhevsky, I. Sutskever, G. E. Hinton, Imagenet classification with deep convolutional neural networks, Communications of the ACM 60 (6) (2017) 84–90

  2. [2]

    Redmon, S

    J. Redmon, S. Divvala, R. Girshick, A. Farhadi, You only look once: Unified, real-time object detection, in: Proceedings of the IEEE confer- ence on computer vision and pattern recognition, 2016, pp. 779–788

  3. [3]

    S. Ren, K. He, R. Girshick, J. Sun, Faster r-cnn: Towards real-time object detection with region proposal networks, Advances in neural in- formation processing systems 28 (2015)

  4. [4]

    C. Finn, P. Abbeel, S. Levine, Model-agnostic meta-learning for fast adaptation of deep networks, in: International conference on machine learning, PMLR, 2017, pp. 1126–1135

  5. [5]

    S. Ravi, H. Larochelle, Optimization as a model for few-shot learning, in: International conference on learning representations, 2017. 26

  6. [6]

    K. Lee, S. Maji, A. Ravichandran, S. Soatto, Meta-learning with differ- entiable convex optimization, in: Proceedings of the IEEE/CVF confer- ence on computer vision and pattern recognition, 2019, pp. 10657–10665

  7. [7]

    Snell, K

    J. Snell, K. Swersky, R. Zemel, Prototypical networks for few-shot learn- ing, Advances in neural information processing systems 30 (2017)

  8. [8]

    Vinyals, C

    O. Vinyals, C. Blundell, T. Lillicrap, D. Wierstra, et al., Matching net- works for one shot learning, Advances in neural information processing systems 29 (2016)

Show all 55 references
  1. [9]

    Zhang, Y

    C. Zhang, Y. Cai, G. Lin, C. Shen, Deepemd: Few-shot image classifica- tion with differentiable earth mover’s distance and structured classifiers, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 12203–12213

  2. [10]

    H. Zhu, R. Zhao, Z. Gao, Q. Tang, W. Jiang, Light transformer learn- ing embedding for few-shot classification with task-based enhancement, Applied Intelligence 53 (7) (2023) 7970–7987

  3. [11]

    Z. Chen, Y. Fu, Y.-X. Wang, L. Ma, W. Liu, M. Hebert, Image de- formation meta-networks for one-shot learning, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 8680–8689

  4. [12]

    Z. Chen, Y. Fu, Y. Zhang, Y.-G. Jiang, X. Xue, L. Sigal, Multi-level semantic feature augmentation for one-shot learning, IEEE Transactions on Image Processing 28 (9) (2019) 4594–4605

  5. [13]

    K. Li, Y. Zhang, K. Li, Y. Fu, Adversarial feature hallucination networks for few-shot learning, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 13470–13479

  6. [14]

    Zhang, S

    Y. Zhang, S. Huang, X. Peng, D. Yang, Semi-identical twins variational autoencoder for few-shot learning, IEEE Transactions on Neural Net- works and Learning Systems (2023)

  7. [15]

    G. S. Dhillon, P. Chaudhari, A. Ravichandran, S. Soatto, A baseline for few-shot image classification, arXiv preprint arXiv:1909.02729 (2019). 27

  8. [16]

    Chen, Y.-C

    W.-Y. Chen, Y.-C. Liu, Z. Kira, Y.-C. F. Wang, J.-B. Huang, A closer look at few-shot classification, in: International Conference on Learning Representations

  9. [17]

    Y. Tian, Y. Wang, D. Krishnan, J. B. Tenenbaum, P. Isola, Rethinking few-shot image classification: a good embedding is all you need?, in: Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XIV 16, Springer, 2020, pp. 266–282

  10. [18]

    M. N. Rizve, S. Khan, F. S. Khan, M. Shah, Exploring complementary strengths of invariant and equivariant representations for few-shot learn- ing, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 10836–10846

  11. [19]

    J. Ma, H. Xie, G. Han, S.-F. Chang, A. Galstyan, W. Abd-Almageed, Partner-assisted learning for few-shot image classification, in: Proceed- ings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 10573–10582

  12. [20]

    Zhang, S

    Y. Zhang, S. Huang, F. Zhou, Generally boosting few-shot learning with handcrafted features, in: Proceedings of the 29th ACM International Conference on Multimedia, 2021, pp. 3143–3152

  13. [21]

    Ouali, C

    Y. Ouali, C. Hudelot, M. Tami, Spatial contrastive learning for few- shot classification, in: Machine Learning and Knowledge Discovery in Databases. Research Track: European Conference, ECML PKDD 2021, Bilbao, Spain, September 13–17, 2021, Proceedings, Part I 21, Springer, 202...

  14. [22]

    H. Lee, S. J. Hwang, J. Shin, Self-supervised label augmentation via in- put transformations, in: International Conference on Machine Learning, PMLR, 2020, pp. 5714–5724

  15. [23]

    Zhang, J

    M. Zhang, J. Zhang, Z. Lu, T. Xiang, M. Ding, S. Huang, Iept: Instance- level and episode-level pretext tasks for few-shot learning, in: Interna- tional Conference on Learning Representations, 2021

  16. [24]

    Y. Rong, X. Lu, Z. Sun, Y. Chen, S. Xiong, Espt: A self-supervised episodic spatial pretext task for improving few-shot learning, in: Pro- ceedings of the AAAI Conference on Artificial Intelligence, 2023. 28

  17. [25]

    Khosla, P

    P. Khosla, P. Teterwak, C. Wang, A. Sarna, Y. Tian, P. Isola, A. Maschinot, C. Liu, D. Krishnan, Supervised contrastive learning, Advances in neural information processing systems 33 (2020) 18661– 18673

  18. [26]

    T. Chen, S. Kornblith, M. Norouzi, G. Hinton, A simple framework for contrastive learning of visual representations, in: International confer- ence on machine learning, PMLR, 2020, pp. 1597–1607

  19. [27]

    K. He, H. Fan, Y. Wu, S. Xie, 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

  20. [28]

    M. Ren, E. Triantafillou, S. Ravi, J. Snell, K. Swersky, J. B. Tenenbaum, H. Larochelle, R. S. Zemel, Meta-learning for semi-supervised few-shot classification, arXiv preprint arXiv:1803.00676 (2018)

  21. [29]

    Bertinetto, J

    L. Bertinetto, J. F. Henriques, P. H. Torr, A. Vedaldi, Meta-learning with differentiable closed-form solvers, arXiv preprint arXiv:1805.08136 (2018)

  22. [30]

    C. Wah, S. Branson, P. Welinder, P. Perona, S. Belongie, The caltech- ucsd birds-200-2011 dataset (2011)

  23. [31]

    H. Guo, L. Shi, Contrastive learning with semantic consistency con- straint, Image and Vision Computing (2023) 104754

  24. [32]

    Z. Wang, S. Shi, Z. Zhai, Y. Wu, R. Yang, Arco: Attention-reinforced transformer with contrastive learning for image captioning, Image and Vision Computing 128 (2022) 104570

  25. [33]

    Y. Zhao, Q. Shu, X. Shi, Dual-level contrastive learning for unsuper- vised person re-identification, Image and Vision Computing 129 (2023) 104607

  26. [34]

    Y. Zhao, Q. Shu, X. Shi, J. Zhan, Unsupervised person re-identification by dynamic hybrid contrastive learning, Image and Vision Computing 137 (2023) 104786. 29

  27. [35]

    D. Chen, Y. Chen, Y. Li, F. Mao, Y. He, H. Xue, Self-supervised learning for few-shot image classification, in: ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, 2021, pp. 1745–1749

  28. [36]

    D. M. Endres, J. E. Schindelin, A new metric for probability distribu- tions, IEEE Transactions on Information theory 49 (7) (2003) 1858– 1860

  29. [37]

    Fuglede, F

    B. Fuglede, F. Topsoe, Jensen-shannon divergence and hilbert space embedding, in: International symposium onInformation theory, 2004. ISIT 2004. Proceedings., IEEE, 2004, p. 31

  30. [38]

    Z. Wu, Y. Xiong, S. X. Yu, D. Lin, Unsupervised feature learning via non-parametric instance discrimination, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 3733– 3742

  31. [39]

    D. Kang, H. Kwon, J. Min, M. Cho, Relational embedding for few-shot classification, in: Proceedings of the IEEE/CVF International Confer- ence on Computer Vision, 2021, pp. 8822–8833

  32. [40]

    Afrasiyabi, J.-F

    A. Afrasiyabi, J.-F. Lalonde, C. Gagn´ e, Associative alignment for few- shot image classification, in: Computer Vision–ECCV 2020: 16th Euro- pean Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part V 16, Springer, 2020, pp. 18–35

  33. [41]

    Mazumder, P

    P. Mazumder, P. Singh, V. P. Namboodiri, Gifsl-grafting based improved few-shot learning, Image and Vision Computing 104 (2020) 104006

  34. [42]

    N. Fei, Z. Lu, T. Xiang, S. Huang, Melr: Meta-learning via modeling episode-level relationships for few-shot learning, in: International Con- ference on Learning Representations, 2021

  35. [43]

    G. Liu, L. Zhao, X. Fang, Pda: Proxy-based domain adaptation for few-shot image recognition, Image and Vision Computing 110 (2021) 104164

  36. [44]

    T. Yu, S. He, Y.-Z. Song, T. Xiang, Hybrid graph neural networks for few-shot learning, in: Proceedings of the AAAI conference on artificial intelligence, Vol. 36, 2022, pp. 3179–3187. 30

  37. [45]

    R. Ma, P. Fang, T. Drummond, M. Harandi, Adaptive poincar´ e point to set distance for few-shot classification, in: Proceedings of the AAAI conference on artificial intelligence, Vol. 36, 2022, pp. 1926–1934

  38. [46]

    Z. Cui, N. Lu, W. Wang, G. Guo, Dual global-aware propagation for few-shot learning, Image and Vision Computing 128 (2022) 104574

  39. [47]

    Zhang, F

    L. Zhang, F. Zhou, W. Wei, Y. Zhang, Meta-hallucinating prototype for few-shot learning promotion, Pattern Recognition 136 (2023) 109235

  40. [48]

    Huang, S

    X. Huang, S. H. Choi, Sapenet: Self-attention based prototype enhance- ment network for few-shot learning, Pattern Recognition 135 (2023) 109170

  41. [49]

    Cheng, Y

    H. Cheng, Y. Wang, H. Li, A. C. Kot, B. Wen, Disentangled feature representation for few-shot image classification, IEEE Transactions on Neural Networks and Learning Systems (2023)

  42. [50]

    R. Gao, H. Su, S. Prasad, P. Tang, Few-shot classification with multi- semantic information fusion network, Image and Vision Computing 141 (2024) 104869

  43. [51]

    Zhang, C

    B. Zhang, C. Luo, D. Yu, X. Li, H. Lin, Y. Ye, B. Zhang, Metadiff: Meta- learning with conditional diffusion for few-shot learning, in: Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 38, 2024, pp. 16687–16695

  44. [52]

    M. Fu, K. Zhu, Instance-based max-margin for practical few-shot recog- nition, in: Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, 2024, pp. 28674–28683

  45. [53]

    H.-J. Ye, H. Hu, D.-C. Zhan, F. Sha, Few-shot learning via embedding adaptation with set-to-set functions, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 8808– 8817

  46. [54]

    Y. Chen, Z. Liu, H. Xu, T. Darrell, X. Wang, Meta-baseline: Explor- ing simple meta-learning for few-shot learning, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 9062–9071. 31

  47. [55]

    W. Xu, Y. Xu, H. Wang, Z. Tu, Attentional constellation nets for few- shot learning, in: International Conference on Learning Representations, 2021. 32

Pith tools

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