Pith. sign in

REVIEW 4 major objections 7 minor 26 references

Compressing Model with Few Class-Imbalance Samples: An Out-of-Distribution Expedition

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

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

Pith's one-line read Out-of-distribution images with complementary labels restore accuracy in few-sample model compression, lifting four existing methods on long-tailed CIFAR-10/100 and ILSVRC-2012.

desk verdict Real problem, plausible empirical gains, but the safety theorem as stated is false and Eq. 6 contradicts its own prose; major revision before it can be trusted. read the letter →

arxiv 2502.05832 v1 pith:X7YBHLIB submitted 2025-02-09 cs.LG cs.AIcs.CV

Tian-Shuang Wu , Shen-Huan Lyu , Ning Chen , Zhihao Qu , Baoliu Ye This is my paper · ORCID
classification cs.LGcs.AIcs.CV
keywords few-shotlearningnetworkcompressionclassimbalanceknowledgedistillationout-of-distributiondatamodelpruninglong-taileddistributionimageclassification
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

Few-sample model compression—turning a large pretrained network into a small deployable one using only a handful of images—is attractive where privacy rules out sharing a full dataset. The paper argues that existing methods were validated on artificially balanced splits (every class equally sampled) and therefore missed a practical failure mode: when the few available samples are long-tailed, minority-class filters get pruned, distillation drifts toward majority classes, and fine-tuning cannot bring the lost features back. It introduces OE-FSMC, a plug-in framework that adds readily available out-of-distribution (OOD) images, labels them with a complementary distribution that rebalances the class prior, and guards against OOD overfitting with a joint distillation loss and a class-dependent regularizer. Wrapped around four existing methods (CD, FSKD, MiR, and PRACTISE) on long-tailed CIFAR-10/100 and ILSVRC-2012, OE-FSMC improves top-1 accuracy, with the largest gains at the smallest sample sizes. If the paper is right, privacy-sensitive deployments can fix imbalance without collecting more in-distribution samples.

What carries the argument

The load-bearing object is the complementary label distribution, $\Gamma_j = (\alpha - \beta_j)/(K\alpha - 1)$, which converts the imbalance in the few-sample set into a rebalanced prior for the mixed dataset. Around it the framework adds three mechanisms: class-aware pruning weights $w_j = p_j/\sum_k p_k$, which rescale filter importance so minority-class channels are not preferentially removed; the joint distillation loss $L = \lambda L_{\mathrm{KD,Few}} + (1-\lambda)L_{\mathrm{KD,OOD}}$, which pools knowledge from real and OOD data while letting $\lambda$ control the balance; and the fine-tuning regularizer $L_{\mathrm{total}} = \mathbb{E}[\ell(f(x;\theta),y)] + \eta \mathbb{E}_{\tilde x \sim P_{\mathrm{aux}}}[\gamma_{\tilde y}\ell(f(\tilde x;\theta),\tilde y)]$, whose class-dependent weight $\gamma_{\tilde y}$ protects minority classes from OOD overfitting. The theoretical anchor is the cited theorem that OOD augmentation with uniformly sampled labels leaves the Bayesian prediction unchanged, which the paper extends, by assertion rather than proof, to complementary labels and to the compression pipeline.

What would settle it

On the same long-tailed CIFAR-10 splits used in the paper, compare OE-FSMC against a variant that labels the OOD images uniformly at random rather than by Eq. 4; if uniform labeling matches OE-FSMC's accuracy, the complementary-label rule is not the active mechanism, and if OE-FSMC fails to beat the no-OOD baseline on minority-class accuracy, the claimed non-harmfulness extension is falsified. A reader could also measure teacher--student penultimate-layer similarity on minority-class validation images after compression to test the asserted feature-alignment mechanism directly.

Watch

Extended reading notes

Core claim

The paper's central claim is that class imbalance is a first-order cause of accuracy loss in few-sample model compression, not a nuisance that fine-tuning can absorb, and that the loss can be repaired by rebalancing the training distribution with out-of-distribution (OOD) data. Each OOD image receives a label drawn from the complementary sampling rule $\Gamma_j = (\alpha - \beta_j)/(K\alpha - 1)$, where $\beta_j = m_j/M$ is the original class frequency and $\alpha = \max_j \beta_j + \min_j \beta_j$, so the mixed class prior approximates the uniform test prior and minority classes receive more auxiliary samples. During pruning, the channel importance score is reweighted by class-aware weights $w_j = p_j / \sum_k p_k$ to protect minority-class filters; during distillation and fine-tuning, a joint loss $L = \lambda L_{\mathrm{KD,Few}} + (1-\lambda)L_{\mathrm{KD,OOD}}$ and a class-dependent regularization term keep the OOD signal from dominating. Experiments wrapping CD, FSKD, MiR, and PRACTISE in this framework on long-tailed CIFAR-10/100 and ILSVRC-2012 report consistent top-1 accuracy gains, most pronounced at 10--20 training samples. The paper grounds the safety of OOD augmentation in a cited theorem stating that uniformly labeled OOD data does not change a Bayesian classifier's prediction, and asserts that this alignment carries over to complementary labels and to the compression and fine-tuning pipeline.

Load-bearing premise

The framework's safety rests on the claim that a theorem proven for OOD data with uniformly sampled labels also holds when labels are drawn from the complementary distribution of Eq. 4 and when the data flows through pruning, distillation, and fine-tuning; the paper asserts this extension without proof, and if the OOD data instead pulls the student model away from the true in-distribution classes, the accuracy gains would reverse.

Editorial extensions

If this is right

  • The four existing compression methods can adopt OE-FSMC without changing their network architecture or reported hyperparameters, making the fix additive rather than a redesign.
  • The accuracy gain is largest when the training set has only 10--20 samples, so the method is most useful exactly where imbalance is hardest to correct manually.
  • An OOD pool of about 500 pre-sampled images, with the auxiliary set roughly the size of the few-sample set, suffices to capture the benefit, easing storage pressure on edge devices.
  • Both integration points contribute to the gain: OOD data during compression preserves minority-class filters that fine-tuning cannot later restore, whereas OOD data during fine-tuning rebalances the classifier; dropping either component lowers accuracy.
  • The authors position the framework as structure-agnostic and point to model quantization as the natural next compression family to try.

Reading between the lines

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

  • Because OE-FSMC acts on the data distribution rather than on the loss landscape, it should compose with algorithmic imbalance remedies such as logit adjustment or class-balanced reweighting; stacking them might close more of the gap to an oracle with real minority-class samples.
  • A testable boundary condition is the distance between the OOD pool and the in-distribution domain: the complementary-label argument predicts graceful degradation only while OOD images carry no information about true classes, so measuring accuracy against OOD-pool similarity would map the safety envelope.
  • The framework's success implies a practical protocol for privacy-preserving deployment: local institutions can simply use whatever images they have on hand as a balancing resource, rather than being asked to collect rare-class examples they cannot obtain.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper addresses class imbalance in few-sample model compression, a practically important setting where only a small imbalanced training set is available. The authors first demonstrate empirically that existing few-sample compression methods (CD, FSKD, MiR, PRACTISE) lose accuracy under long-tailed imbalance. They then propose OE-FSMC, a framework that adds out-of-distribution (OOD) data with complementary labels during both the compression and fine-tuning stages, together with a joint distillation loss and a class-dependent regularization term. The experiments on long-tailed CIFAR-10/100 and ILSVRC-2012 show consistent accuracy improvements when OE-FSMC is combined with each of the four base methods. The paper claims a theoretical guarantee, based on Theorem 1 of Wei et al. (2022), that the complementary sampling strategy is harmless to the Bayesian classifier's prediction.

Significance. If the empirical claims hold, the paper addresses a real and understudied problem: few-sample model compression under class imbalance. The framework is simple, compatible with several existing compression pipelines, and shows consistent gains, especially at very small sample sizes. The experimental coverage is reasonably broad (three datasets, four base methods, ablations, hyperparameter sensitivity). The main value is empirical; however, the theoretical justification as written is invalid, and one component of the method (Eq. 6) contradicts its stated purpose. These issues must be fixed before the paper can be considered sound, but the core empirical idea may be salvageable after a substantial revision of the method description and theoretical claims.

major comments (4)
  1. [§4.2, Eq. (4) and Theorem 1] The complementary label distribution Γ_j = (α - β_j)/(K·α - 1) is non-uniform whenever the class priors β are imbalanced, yet Theorem 1 is stated only for labels uniformly sampled from the label space. The paper provides no proof that Eq. (4) inherits the "harmless" property. In fact, the claim is false for a natural counterexample: with K=2, β=(0.9, 0.1), α=1, so Γ=(0.1, 0.9), and if the OOD support is disjoint from the in-distribution support (as the paper's OOD purity step implies), then for any in-distribution x the mixed posterior satisfies P_mix(x|y) ∝ P_s(x|y) while P_mix(y) = (1-π)β_y + πΓ_y. For any π > 0.5, argmax_y P_mix(x|y)P_mix(y) flips to the minority class for points where the original Bayes classifier chose the majority class. This directly contradicts the assertion that Eq. (4) "ensures the stability of the Bayesian classifier's prediction" and invalidates the theoretical basis for the framework's central safety claim. The authors must either prove a correct invariance condition for their sampling scheme, adopt a label distribution for which the theorem actually applies (e.g., uniform) and find another way to rebalance, or substantially reframe the method as a purely empirical heuristic without the "non-harmful" guarantee.
  2. [§4.3, Eq. (6)] The class-aware weight w_j = p_j / Σ_k p_k, with p_j the relative frequency of class j, assigns smaller weights to minority classes, directly contradicting the prose that "Minority classes are assigned higher weights through this formula." As written, Eqs. (5)-(6) would further downweight minority classes in the channel importance score, exacerbating the very imbalance problem the paper aims to solve. This is a load-bearing inconsistency in the method definition; it must be corrected (for example, by using an inverse-frequency weighting such as w_j = (1/p_j) / Σ_k (1/p_k)) and the surrounding text must be made consistent with the formula actually used in the experiments.
  3. [§4.2, inference from Bayes classifier to student feature alignment] Even if the Bayesian classifier invariance held, the statement "we can infer that the feature space of the student model remains aligned with that of the teacher model on Dmix" is a non-sequitur: Theorem 1 concerns the Bayes-optimal decision over the input space, not the intermediate feature alignment produced by distillation or pruning. The paper should either provide an argument connecting the two, or explicitly state this as an assumption that is not proven. As it stands, the theoretical pipeline from Theorem 1 to the method's design is incomplete.
  4. [§4.1, Eq. (1)] The analysis of pruning bias assumes that the importance score of a filter is a linear function of class frequencies, s_k^{(l)} = Σ_j (m_j/M) s_{k,j}^{(l)}. This is an illustrative assumption, but it is not generally true for the pruning criteria used by the evaluated methods (e.g., PRACTISE uses block dropping with a recoverability metric, not a class-frequency-weighted filter score). The theoretical motivation for why class imbalance harms compression should be reconciled with the actual mechanisms of the evaluated methods, or the analysis should be explicitly labeled as a motivating heuristic rather than a general derivation.
minor comments (7)
  1. [§4.2-§4.3] The symbol γ is used both for the complementary sampling rate in Eq. (4) and for the class-dependent regularization weight in Eqs. (8)-(9); using distinct symbols (e.g., Γ_j for the sampling rate and γ_y for the weight) would prevent confusion.
  2. [§4.1] The text switches to first-person singular in several places ("I will analyze", "I propose") despite multiple authors; the manuscript should use "we" consistently.
  3. [Figure 3] In the version provided, the text in Figure 3 is corrupted (sequences like "/uni00000026/uni00000027/..." appear instead of readable labels). The authors must ensure the final PDF renders all figure text correctly; as it stands, the figure is unreadable and the ablation discussion in §5.4 cannot be fully verified.
  4. [§5.4] The ablation study is reported only through the unreadable Figure 3(a) and qualitative statements; no numerical results, error bars, or per-component accuracy values are given, making it hard to assess the magnitude of each component's contribution.
  5. [Table 2] The "Num" row mixes different total sample sizes across datasets (10/20/50/100 for CIFAR-10, 200/500/1000 for CIFAR-100, 1000/2000/3000 for ILSVRC-2012). The caption should clarify which numbers correspond to which dataset and that the ILSVRC-2012 settings correspond to 1/2/3 samples per class.
  6. [§5.2] The explanation that PRACTISE is less affected by imbalance "likely because the block-dropping strategy is less sensitive to imbalance" is speculative; a per-class accuracy breakdown or a latency-accuracy comparison under imbalance would substantiate this claim.
  7. [References] Several reference entries have formatting issues (e.g., "32rd" in Zhou et al. 2020) and some entries are missing page numbers; a careful proofread of the bibliography is needed.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: OE-FSMC's gains are measured against external baselines, and the only self-citation is a non-load-bearing related-work reference.

full rationale

The claimed derivation chain for OE-FSMC does not reduce to its own inputs. The method consists of design choices: complementary-label OOD assignment (Eq. 4), a weighted channel-importance score (Eqs. 5-6), a joint distillation loss (Eq. 7), and a regularized fine-tuning objective (Eqs. 8-9). These are constructions whose value is tested empirically in Tables 1-2 against the unmodified baselines CD, FSKD, MiR, and PRACTISE, with the original hyperparameters retained. The reported accuracy improvements are therefore external benchmark outcomes, not predictions forced by fitted parameters. The only theoretical ingredient, Theorem 1, is imported from Wei et al. (2022), an external paper. Section 4.2 explicitly quotes the theorem's uniform-label condition, then extends it to the complementary distribution of Eq. 4 without proof. That extension is a gap in justification rather than circularity, because the theorem is not derived from the present method and the paper does not define the complementary distribution in terms of the theorem's conclusion. The one self-citation, He et al. (2024) in Section 2.2, is co-authored by S.-H. Lyu, but it is a related-work reference used only to catalog algorithm-level imbalance methods; it is not load-bearing for any of the paper's claims. Thus the paper is self-contained against external benchmarks, and the appropriate circularity finding is no significant circularity; the minor self-citation accounts for the score of 2.

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

The method introduces no new physical entities. The load-bearing pieces pulled from outside the paper's own derivation are the Open Sampling theorem and several hand-chosen weighting hyperparameters, catalogued above.

free parameters (5)
  • lambda (distillation balance) = 0.5
    Tuned via sensitivity analysis (Figure 3c) using test accuracy; no held-out validation set.
  • eta (regularization weight) = 2.5
    Tuned via sensitivity analysis (Figure 3d) using test accuracy.
  • gamma_y (class-dependent weight) = undefined
    Appears in Eqs. 8 and 9 but is never defined or specified; load-bearing for the fine-tuning regularization.
  • w_j (class-aware pruning weight) = p_j / sum(p_k), inconsistent with claimed purpose
    As written, minority classes with small p_j receive small weights, not the higher weights the text claims; a likely typo, but it is central to the pruning adjustment.
  • OOD auxiliary set size = matched to few-shot set size; 500 pre-sampled
    Section 4.3 and Figure 3b show the best size is comparable to the training set, so the value is chosen empirically rather than derived.
assumptions (4)
  • domain assumption Theorem 1 of Wei et al. (2022): adding OOD data with uniformly sampled labels does not change the Bayesian classifier's argmax.
    Invoked in Section 4.2 but not proved; the paper extends it to a compression setting and to non-uniform complementary labels without proof.
  • ad hoc to paper The complementary label distribution (Eq. 4) preserves the student-teacher feature alignment during compression.
    Stated in Section 4.2 as a consequence of Theorem 1, but Theorem 1 only covers uniform labels; no separate proof or experiment is given.
  • ad hoc to paper Pruning importance scores are linear in class frequencies (Eq. 1), so minority classes lose channels.
    Presented in Section 4.1 as the motivation for the class-aware weights; it is an assumed form of the importance score, not demonstrated for the actual pruning methods (CD, FSKD, MiR, PRACTISE).
  • domain assumption A 500-image random sample of TinyImage, after overlap removal, is a pure and sufficient OOD source.
    Section 4.3 states this without analysis of label purity or distribution shift; the OOD source is not shipped or precisely specified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Compressing Model with Few Class-Imbalance Samples: An Out-of-Distribution Expedition." pith.science (2026). https://pith.science/paper/X7YBHLIB

@misc{pith2026250205832,
  author       = {Pith},
  title        = {Pith review of: Compressing Model with Few Class-Imbalance Samples: An Out-of-Distribution Expedition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/X7YBHLIB}},
  note         = {Machine review of arXiv:2502.05832}
}
read the original abstract

In recent years, as a compromise between privacy and performance, few-sample model compression has been widely adopted to deal with limited data resulting from privacy and security concerns. However, when the number of available samples is extremely limited, class imbalance becomes a common and tricky problem. Achieving an equal number of samples across all classes is often costly and impractical in real-world applications, and previous studies on few-sample model compression have mostly ignored this significant issue. Our experiments comprehensively demonstrate that class imbalance negatively affects the overall performance of few-sample model compression methods. To address this problem, we propose a novel and adaptive framework named OOD-Enhanced Few-Sample Model Compression (OE-FSMC). This framework integrates easily accessible out-of-distribution (OOD) data into both the compression and fine-tuning processes, effectively rebalancing the training distribution. We also incorporate a joint distillation loss and a regularization term to reduce the risk of the model overfitting to the OOD data. Extensive experiments on multiple benchmark datasets show that our framework can be seamlessly incorporated into existing few-sample model compression methods, effectively mitigating the accuracy degradation caused by class imbalance.

Figures

Figures reproduced from arXiv: 2502.05832 by the authors.

Figure 1
Figure 1. Overview of introduction. Left: Illustration of few-sample model compression workflow for deployment in small hospitals. Right: Class imbalance in disease cases. 1 Introduction As deep learning technology has advanced, models have become larger and more complex, resulting in challenges related to computational resources and storage in real-world appli￾cations. For instance, CNNs with millions of parameters cannot be… view at source ↗
Figure 2
Figure 2. Illustration of OE-FSMC. Left: Label assignment strategy for OOD data. Right: Framework of OE-FSMC in Compression and Fine-Tuning Process. s (l) k is defined as: s (l) k = X K j=1 mj M s (l) k,j , (1) where s (l) k,j represents the importance contribution of filter k to class j. In the case of class imbalance, the term mj M is small for the minority class, which weakens its contribution. This leads to the erroneous … view at source ↗
Figure 3
Figure 3. Results of extra experiments.In this figure: a) Results of ablation study. b) Effect of different sizes of the auxiliary set. c) Sensitivity analysis of λ ∈ [0, 1]. d) Sensitivity analysis of η ∈ [0.1, 5]. our method are both highly effective in mitigating the detrimental effects of class imbalance in the few-sample compression process. Using either component alone improves the perfor￾mance of the compressed model. … view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 25 canonical work pages

  1. [1]

    Abdi and S

    L. Abdi and S. Hashemi. To combat multi-class imbalanced problems by means of over-sampling techniques. IEEE Transactions on Knowledge and Data Engineering, 28 0 (1): 0 238--251, 2015

    work page 2015

  2. [2]

    H. Bai, J. Wu, I. King, and M. Lyu. Few shot network compression via cross distillation. In Proceedings of the 34th AAAI Conference on Artificial Intelligence, pages 3203--3210, 2020

    work page 2020

  3. [3]

    N. V. Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer. Smote: S ynthetic minority over-sampling technique. Journal of Artificial Intelligence Research, 16: 0 321--357, 2002

    work page 2002

  4. [4]

    N. V. Chawla, A. Lazarevic, L. O. Hall, and K. W. Bowyer. Smoteboost: Improving prediction of the minority class in boosting. In Proceedings of 7th European Conference on Principles and Practice of Knowledge Discovery in Databases, pages 107--119, 2003

    work page 2003

  5. [5]

    X. Dong, S. Chen, and S. Pan. Learning to prune deep neural networks via layer-wise optimal brain surgeon. Advances in Neural Information Processing Systems, 30, 2017

    work page 2017

  6. [6]

    Fern \'a ndez, S

    A. Fern \'a ndez, S. Garc \' a, M. Galar, R. C. Prati, B. Krawczyk, F. Herrera, A. Fern \'a ndez, S. Garc \' a, M. Galar, R. C. Prati, et al. Cost-sensitive learning. In Learning from Imbalanced Data Sets, pages 63--78. Springer, Cham, 2018

    work page 2018

  7. [7]

    Galar, A

    M. Galar, A. Fernandez, E. Barrenechea, H. Bustince, and F. Herrera. A review on ensembles for the class imbalance problem: bagging-, boosting-, and hybrid-based approaches. IEEE Transactions on Systems, Man, and Cybernetics, Part C, 42 0 (4): 0 463--484, 2011

    work page 2011

  8. [8]

    Y. He, X. Zhang, and J. Sun. Channel pruning for accelerating very deep neural networks. In Proceedings of the 16th the IEEE International Conference on Computer Vision, pages 1398--1406, 2017

    work page 2017

Show all 26 references
  1. [9]

    He, D.-X

    Y.-X. He, D.-X. Liu, S.-H. Lyu, C. Qian, and Z.-H. Zhou. Multi-class imbalance problem: A multi-objective solution. Information Sciences, 680: 0 121156, 2024

    2024

  2. [10]

    G. Hinton. Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531, 2015

    2015 arXiv

  3. [11]

    Japkowicz and S

    N. Japkowicz and S. Stephen. The class imbalance problem: A systematic study. Intelligent Data Analysis, 6 0 (5): 0 429--449, 2002

    2002

  4. [12]

    Krizhevsky and G

    A. Krizhevsky and G. Hinton. Learning multiple layers of features from tiny images. Technical Report UT-ML-TR-2009-001, University of Toronto, 2009

    2009

  5. [13]

    H. Li, A. Kadav, I. Durdanovic, H. Samet, and H. P. Graf. Pruning filters for efficient convnets. In Proceedings of the 5th International Conference on Learning Representations, 2017

    2017

  6. [14]

    T. Li, J. Li, Z. Liu, and C. Zhang. Few sample knowledge distillation for efficient network compression. In Proceedings of the 32nd IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14639--14647, 2020

    2020

  7. [15]

    Lin, C.-F

    W.-C. Lin, C.-F. Tsai, Y.-H. Hu, and J.-S. Jhang. Clustering-based undersampling in class-imbalanced data. Information Sciences, 409: 0 17--26, 2017

    2017

  8. [16]

    X.-Y. Liu, J. Wu, and Z.-H. Zhou. Exploratory undersampling for class-imbalance learning. IEEE Transactions on Systems, Man, and Cybernetics, Part B, 39 0 (2): 0 539--550, 2008

    2008

  9. [17]

    Mohammed, J

    R. Mohammed, J. Rawashdeh, and M. Abdullah. Machine learning with oversampling and undersampling techniques: O verview study and experimental results. In Proceedings of the 11th International Conference on Information and Communication Systems, pages 243--248, 2020

    2020

  10. [18]

    Nagel, M

    M. Nagel, M. v. Baalen, T. Blankevoort, and M. Welling. Data-free quantization through weight equalization and bias correction. In Proceedings of the 17th IEEE/CVF International Conference on Computer Vision, pages 1325--1334, 2019

    2019

  11. [19]

    Ochal, M

    M. Ochal, M. Patacchiola, J. Vazquez, A. Storkey, and S. Wang. Few-shot learning with class imbalance. IEEE Transactions on Artificial Intelligence, 2023

    2023

  12. [20]

    Romero, N

    A. Romero, N. Ballas, S. E. Kahou, A. Chassang, C. Gatta, and Y. Bengio. Fitnets: Hints for thin deep nets. In Proceedings of 3rd International Conference on Learning Representations, pages 1--13, 2015

    2015

  13. [21]

    Sharma, A

    S. Sharma, A. Gosain, and S. Jain. A review of the oversampling techniques in class imbalance problem. In Proceedings of the 5th International Conference on Innovative Computing and Communications, pages 459--472, 2022

    2022

  14. [22]

    Wang and J

    G.-H. Wang and J. Wu. Practical network acceleration with tiny sets. In Proceedings of the 36th IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20331--20340, 2023

    2023

  15. [23]

    H. Wang, J. Liu, X. Ma, Y. Yong, Z. Chai, and J. Wu. Compressing models with few samples: M imicking then replacing. In Proceedings of the 35th IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 701--710, 2022

    2022

  16. [24]

    H. Wei, L. Tao, R. Xie, L. Feng, and B. An. Open-sampling: Exploring out-of-distribution data for re-balancing long-tailed datasets. In Proceedings of the 39th International Conference on Machine Learning, pages 23615--23630, 2022

    2022

  17. [25]

    B. Zhou, Q. Cui, X.-S. Wei, and Z.-M. Chen. Bbn: Bilateral-branch network with cumulative learning for long-tailed visual recognition. In Proceedings of the 32rd IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9719--9728, 2020

    2020

  18. [26]

    Zhou and X.-Y

    Z.-H. Zhou and X.-Y. Liu. Training cost-sensitive neural networks with methods addressing the class imbalance problem. IEEE Transactions on Knowledge and Data Engineering, 18 0 (1): 0 63--77, 2005

    2005

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.