Pith. sign in

REVIEW 3 major objections 5 minor 40 references

Improving Noise Efficiency in Privacy-preserving Dataset Distillation

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Dosser, a framework for differentially private dataset distillation, achieves a 10.0% accuracy gain on CIFAR-10 at 50 images per class and an 8.3% gain at one-fifth the distilled set size by decoupling sampling from optimization and…

desk verdict DOS is a solid empirical increment, but Lemma 1's sensitivity reduction claim is wrong and the DP accounting needs fixing before the paper can be trusted. read the letter →

arxiv 2508.01749 v1 pith:DUVZEY7B submitted 2025-08-03 cs.CV cs.AI

Runkai Zheng , Vishnu Asutosh Dasu , Yinong Oliver Wang , Haohan Wang , Fernando De la Torre This is my paper
classification cs.CVcs.AI
keywords differentialprivacydatasetdistillationsyntheticdatagenerationsubspaceprojectionsignal-to-noiseratiomatching-baseddecoupledoptimizationprivacy-preservingmachinelearning
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 tries to show that the main waste in privacy-preserving dataset distillation is not the privacy budget itself but how the budget is spent: existing matching-based methods add fresh differentially private noise at every sampling step and match raw signals from randomly initialized networks, so much of the added noise carries little information. The proposed framework, Dosser, separates the sampling of private training signals from the optimization of synthetic images, and projects signals into a low-dimensional informative subspace before noise is added. With the same privacy budget, this yields a cleaner training signal and a much smaller distilled dataset. If the claim is correct, private synthetic data becomes more practical: more accurate compact datasets under a fixed privacy guarantee, or the same accuracy under a tighter one.

What carries the argument

The load-bearing mechanism has two parts. DOS (Decoupled Optimization and Sampling) first samples and stores $N_s$ DP-protected aggregated training signals, one per iteration, and then runs $N_o \gg N_s$ optimization steps that reuse those stored signals, so extra convergence costs no extra privacy. SER (Subspace-based Error Reduction) learns a projection matrix $P$ by PCA on an auxiliary dataset and applies it before clipping and Gaussian noise, so matching happens in a $k$-dimensional subspace; the paper's Theorem 1 decomposes the mean-squared error of mean estimation into projection residual, dimensional-reduction effect, and projection error to show when projection helps. The optimization loss is the squared $\ell^2$ distance between synthetic and noisy real signals in the projected space, and the stored tuples also carry the augmentation seed and random network initialization used to extract each signal.

What would settle it

Compute the $\ell^2$ sensitivity of the projected mean function $f_{\text{proj}}(D)=\tfrac{1}{n}\sum_i P\phi(x_i)$ under the add/remove-one adjacent-dataset model; if it equals the original sensitivity instead of the claimed smaller value, Lemma 1's noise-scale reduction fails. A simple control experiment would replace SER's PCA projection with a random orthonormal projection of the same dimension: if accuracy stays the same, the gain is dimensionality reduction rather than the learned informative directions.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that noise efficiency, not raw signal quantity, is what decides how much accuracy private dataset distillation can keep. For matching-based distillation under $(10,10^{-5})$-DP, Dosser reports a 10.0% accuracy improvement on CIFAR-10 with 50 images per class and an 8.3% increase with just one-fifth the distilled set size of previous state-of-the-art methods, while staying close to the non-private distribution-matching baseline. The gain comes from two changes: DOS decouples the sampling of DP-protected signals from the optimization of synthetic images so that many optimization steps can reuse a fixed set of privatized signals, and SER projects those signals into a subspace learned from auxiliary data, concentrating informative structure before Gaussian noise is added. The auxiliary data can come from a public generator at no privacy cost, or from a DP generative model trained under a split budget, with the total guarantee preserved by composition.

Load-bearing premise

The load-bearing premise is that projecting training signals into a lower-dimensional subspace reduces the worst-case influence any single person's data can have on the released signal, so less Gaussian noise is needed for the same privacy guarantee; if projection does not actually shrink that worst-case influence, the theoretical component of the noise-saving argument weakens.

Editorial extensions

If this is right

  • Under the paper's results, a fixed DP budget buys more accurate distilled datasets: on CIFAR-10, 50 images per class gain 10.0% over the previous state of the art, and an 8.3% accuracy increase is obtained with one-fifth the distilled set size.
  • Because DOS lets optimization run far longer than sampling, extra compute can improve synthetic-image convergence without spending additional privacy budget.
  • Because SER's auxiliary dataset can come from a public generator, natural-image domains get the subspace benefit at zero additional privacy cost, while specialized domains can pay for auxiliary data out of the same budget through a DP generative model and still keep the overall guarantee by composition.
  • The framework accepts any matching signal, features or gradients, so the same decoupling and subspace projection can be applied to other matching-based private distillation methods.

Reading between the lines

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

  • A likely reason SER helps in practice, even if its sensitivity-reduction argument is set aside, is that DP Gaussian noise is isotropic while signal variance is not: PCA concentrates high-variance signal into few coordinates, so the retained coordinates have a higher signal-to-noise ratio; the paper's own observation that keeping all 2048 dimensions still helps is consistent with variance concentra
  • The same 'sample private signals once, optimize forever' pattern could be lifted into other DP pipelines, such as private fine-tuning or private data generation, whenever a fixed set of privatized gradients or features can support many update steps.
  • A testable prediction from the error decomposition is that SER's benefit should grow as the privacy budget tightens and shrink as the budget loosens, matching the U-shaped curves the paper reports in its appendix.
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

3 major / 5 minor

Summary. The paper proposes Dosser, a framework for differentially private dataset distillation that combines Decoupled Optimization and Sampling (DOS) with Subspace-based Error Reduction (SER). DOS first samples a fixed set of DP-protected training signals and then performs many optimization steps on the synthetic dataset using those stored signals, avoiding the need to add fresh noise at every optimization step. SER projects the extracted training signals onto a low-dimensional subspace learned from an auxiliary dataset (generated by Stable Diffusion for CIFAR-10, or by a DP diffusion model for MNIST/FashionMNIST), with the goal of reducing the impact of DP noise. The paper reports substantial accuracy gains over prior private distillation methods on MNIST, FashionMNIST, and CIFAR-10 under (10,10^-5)-DP, including a 10.0% improvement on CIFAR-10 at IPC=50 and an 8.3% improvement with one-fifth the distilled set size. The theoretical contribution is an MSE decomposition (Theorem 1) that separates projection residual, dimensional-reduction, and projection-error terms, supported by a lemma about the sensitivity of the projected sample mean (Lemma 1) and a composition theorem for the auxiliary-data budget (Theorem 2).

Significance. If the empirical results are correct, the paper would make a useful practical contribution to private dataset distillation: the decoupling idea is simple and effective, and the subspace-projection mechanism is well motivated. The authors provide code, extensive ablations, and comparisons with several baselines, which strengthens the empirical part. However, the core theoretical justification for SER, specifically the claim that projecting the signal reduces the L2 sensitivity and therefore permits a smaller Gaussian noise scale, is flawed. Because differential privacy is the central promise of the paper, this error is load-bearing: the stated privacy guarantee is not established for the SER component as described. The empirical improvements might still be real under a corrected post-processing interpretation, but the current manuscript overstates its theoretical contribution and needs a substantial revision of the privacy analysis.

major comments (3)
  1. [Appendix A, Lemma 1] The sensitivity reduction claimed in Lemma 1 is invalid. For a fixed matrix P with orthonormal rows (P P^T = I_k), the L2 sensitivity of the projected sample mean f(D) = P * (1/n) * sum clip(phi(x_i)) under add/remove adjacency is sup_{||v|| <= B} ||P v|| / n = B/n, because the supremum is attained by v lying in the row space of P (e.g., v = P^T y with ||y|| = B). This equals the original sensitivity, not sqrt(k/d) * B/n. The proof incorrectly computes the norm of a single difference vector rather than the supremum over all adjacent pairs. Consequently, the ratio sigma_orig / sigma_proj derived in the lemma is unjustified, and if the implementation sets sigma_proj = sqrt(k/d) sigma_orig, the Gaussian mechanism with that reduced noise scale does not satisfy the stated (epsilon, delta)-DP guarantee.
  2. [Theorem 1 and Section 3.3] The 'Dimensional Reduction Effect' term in Theorem 1 inherits the error of Lemma 1: it relies on the false claim that the DP noise scale can be reduced by a factor related to the subspace dimension. The MSE decomposition therefore overstates the benefit of subspace projection for reducing DP noise. To make the theorem valid, the authors should either (a) add the noise in the original d-dimensional space and treat projection as post-processing, in which case sigma_proj = sigma_orig and the reconstruction noise variance is reduced because the noise covariance after projection and reconstruction is sigma^2 * P^T P, or (b) provide a correct sensitivity analysis for a mechanism that genuinely adds noise in the projected space. As written, the theorem's second term is unsupported and the claimed noise-efficiency gain is not established.
  3. [Section 3.4 and Algorithm 1] The description of the sampling stage is ambiguous about the order of projection, clipping, and noise addition. The text says the noisy mean in the projected space is obtained by projecting and then adding Gaussian noise with scale sigma_proj, and Section 3.3 explicitly allows sigma_proj to be smaller than sigma_orig. If the implementation follows this description, the privacy guarantee fails. If the implementation instead adds noise in the original space and then projects (which would be valid as post-processing), the text and the MSE analysis must be revised to reflect that sigma_proj = sigma_orig. The paper needs to state the exact mechanism unambiguously and give a correct sensitivity analysis for that mechanism.
minor comments (5)
  1. [Abstract and Section 4.1] Calling (10, 10^-5) a 'strict privacy budget' is misleading, since epsilon=10 is a high privacy parameter (weak privacy guarantee). Consider rephrasing and reporting results for lower epsilon values, e.g., epsilon in {1,2,5}, to support the privacy-preserving claim.
  2. [Theorem 1 equation] The display of Theorem 1 is difficult to parse because several symbols (e.g., the exact form of the 'Dimensional Reduction Effect' term) are not clearly defined in the main text. Please re-typeset the equation with explicit notation and define every quantity used in the decomposition.
  3. [Appendix D.1] The claim that Stable Diffusion pretrained on LAION-5B 'contains no information related to CIFAR-10' is not empirically substantiated. Since CIFAR-10 is a widely used public dataset, web-scraped corpora may contain images from it; a brief discussion of potential distribution overlap would strengthen the privacy argument.
  4. [Appendix B.1, Table 3] The caption of Table 3 should explicitly state the total privacy budget (epsilon=10, delta=10^-5) and clarify that the split satisfies epsilon1 + epsilon2 = 10, so the comparison of budget splits is easy to follow.
  5. [Section 4.2, Table 1] The table reports mean accuracies without confidence intervals or standard deviations; adding these would help assess the statistical significance of the reported improvements, especially for the 8.3% and 10.0% claims.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the empirical accuracy gains are measured against external baselines, and the theoretical MSE decomposition is a derivation from stated assumptions rather than a fit to the target results.

full rationale

The paper's central performance claims are experimental: Dosser's accuracy improvements on MNIST, FashionMNIST, and CIFAR-10 are reported as measured outcomes compared with published baselines, not as predictions derived from the theory. The theoretical content (Theorem 1 and Appendix A) is an algebraic MSE decomposition: given Assumption 1's signal/covariance model, the difference between MSE with and without projection is expanded into projection residual, dimensional-reduction, and projection-error terms. This is a derivation from stated assumptions, not a quantity fitted to the later accuracy numbers. The subspace dimension, auxiliary dataset size, and privacy-budget split are selected empirically or by validation accuracy, which is hyperparameter tuning rather than fitted-input-called-prediction. The use of a DP generative model on MNIST/FashionMNIST to build the auxiliary dataset is explicitly accounted for by composition in Theorem 2, which restates the standard basic-composition and post-processing theorems; it does not smuggle in the result. The only author-overlapping citation is the related-work mention of Wasserstein dataset distillation, which is not load-bearing. The reviewer's concern about Lemma 1's sensitivity computation is a mathematical correctness issue, not a circularity: even if the supremum argument is flawed, the claim is attempted to be proven rather than assumed as the conclusion, so it does not reduce the derivation to its inputs. No self-definitional, fitted-prediction, or self-citation-chain circularity is present.

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

The method introduces no new physical or mathematical entities. Its load-bearing assumptions are the signal decomposition, the alignment of auxiliary and private distributions, and the sensitivity-reduction claim in Lemma 1, the last of which is questionable.

free parameters (5)
  • Subspace dimension k = 500 for CIFAR-10
    Chosen by the user; directly controls the SER trade-off between noise reduction and projection error.
  • Auxiliary dataset size = 1000
    Chosen by the user; the paper finds minimal impact beyond enabling higher PCA dimensions.
  • Privacy budget split (epsilon1, epsilon2) = (0.8, 0.2) for MNIST/FashionMNIST
    Selected empirically on the validation set, as stated in Section 4.1 and Appendix B.1.
  • Sampling iterations T_s and optimization iterations T_o = 10,000 and 200,000
    Chosen hyperparameters; T_o much larger than T_s is the core of DOS.
  • Clipping threshold C
    Used in Eq. (1) but the specific value is not reported in the paper.
assumptions (4)
  • domain assumption Assumption 1: each signal vector decomposes into true mean, informative zero-mean component with covariance Sigma_u, and uninformative zero-mean component with covariance Sigma_n.
    This structural assumption about the signal distribution is the basis for Theorem 1 and is stated in Section 3.3.
  • domain assumption The auxiliary dataset distribution closely aligns with the private data distribution in the signal space.
    SER's benefit depends on the PCA subspace learned from auxiliary data matching the informative subspace of private data; the paper acknowledges this in Limitations.
  • ad hoc to paper Lemma 1: projecting the signal reduces the L2 sensitivity by a multiplicative factor sqrt(k/d), allowing a proportionally smaller noise scale.
    This is a mathematically asserted claim in Appendix A that is not justified for a fixed orthonormal projection and appears to be incorrect.
  • domain assumption Stable Diffusion v1-4, trained on LAION-5B, contains no information related to CIFAR-10, so using it for auxiliary data does not cause privacy leakage.
    Stated in Appendix D.1 without empirical verification; if LAION-5B contains CIFAR-10-like images, the argument weakens.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Improving Noise Efficiency in Privacy-preserving Dataset Distillation." pith.science (2026). https://pith.science/paper/DUVZEY7B

@misc{pith2026250801749,
  author       = {Pith},
  title        = {Pith review of: Improving Noise Efficiency in Privacy-preserving Dataset Distillation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DUVZEY7B}},
  note         = {Machine review of arXiv:2508.01749}
}
read the original abstract

Modern machine learning models heavily rely on large datasets that often include sensitive and private information, raising serious privacy concerns. Differentially private (DP) data generation offers a solution by creating synthetic datasets that limit the leakage of private information within a predefined privacy budget; however, it requires a substantial amount of data to achieve performance comparable to models trained on the original data. To mitigate the significant expense incurred with synthetic data generation, Dataset Distillation (DD) stands out for its remarkable training and storage efficiency. This efficiency is particularly advantageous when integrated with DP mechanisms, curating compact yet informative synthetic datasets without compromising privacy. However, current state-of-the-art private DD methods suffer from a synchronized sampling-optimization process and the dependency on noisy training signals from randomly initialized networks. This results in the inefficient utilization of private information due to the addition of excessive noise. To address these issues, we introduce a novel framework that decouples sampling from optimization for better convergence and improves signal quality by mitigating the impact of DP noise through matching in an informative subspace. On CIFAR-10, our method achieves a \textbf{10.0\%} improvement with 50 images per class and \textbf{8.3\%} increase with just \textbf{one-fifth} the distilled set size of previous state-of-the-art methods, demonstrating significant potential to advance privacy-preserving DD.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

40 extracted references · 33 canonical work pages

  1. [1]

    Deep learning with differential privacy

    Martin Abadi, Andy Chu, Ian Goodfellow, H Brendan McMahan, Ilya Mironov, Kunal Talwar, and Li Zhang. Deep learning with differential privacy. InProceedings of the 2016 ACM SIGSAC conference on computer and communications security, pages 308–318, 2016. 1

    2016

  2. [2]

    Don’t generate me: Training differen- tially private generative models with sinkhorn divergence

    Tianshi Cao, Alex Bie, Arash Vahdat, Sanja Fidler, and Karsten Kreis. Don’t generate me: Training differen- tially private generative models with sinkhorn divergence. Advances in Neural Information Processing Systems , 34: 12480–12492, 2021. 6, 8

    work page 2021

  3. [3]

    Cazenavette

    George et al. Cazenavette. Dataset distillation by matching training trajectories. In CVPR, 2022. 2

    work page 2022

  4. [4]

    Private set generation with discriminative information

    Dingfan Chen, Raouf Kerkouche, and Mario Fritz. Private set generation with discriminative information. Advances in Neural Information Processing Systems, 35:14678–14690,

    work page

  5. [5]

    An analysis of single-layer networks in unsupervised feature learning

    Adam Coates, Andrew Ng, and Honglak Lee. An analysis of single-layer networks in unsupervised feature learning. In AISTATS, pages 215–223. JMLR, 2011. 6

    work page 2011

  6. [6]

    Justin et al. Cui. Scaling up dataset distillation to imagenet- 1k with constant memory. In ICML, 2023. 3

    work page 2023

  7. [7]

    The mnist database of handwritten digit images for machine learning research

    Li Deng. The mnist database of handwritten digit images for machine learning research. IEEE Signal Processing Maga- zine, 29(6):141–142, 2012. 6

    work page 2012

  8. [8]

    Differentially private diffusion models.arXiv preprint arXiv:2210.09929, 2022

    Tim Dockhorn, Tianshi Cao, Arash Vahdat, and Karsten Kreis. Differentially private diffusion models.arXiv preprint arXiv:2210.09929, 2022. 6

    arXiv 2022

Show all 40 references
  1. [9]

    Minimizing the accumulated trajectory error to improve dataset distillation

    Jiawei Du, Yidi Jiang, Wenqing Wang, Zhenyu Qin, and Zheng-Jun Zha. Minimizing the accumulated trajectory error to improve dataset distillation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7059–7068, 2023. 3

    2023

  2. [10]

    The algorithmic founda- tions of differential privacy

    Cynthia Dwork and Aaron Roth. The algorithmic founda- tions of differential privacy. Found. Trends Theor. Comput. Sci., 9(3–4):211–407, 2014. 2, 4, 3

    2014

  3. [11]

    Differentially private diffusion models generate useful synthetic images

    Sahra Ghalebikesabi, Leonard Berrada, Sven Gowal, Ira Ktena, Robert Stanforth, Jamie Hayes, Soham De, Samuel L Smith, Olivia Wiles, and Borja Balle. Differentially private diffusion models generate useful synthetic images. arXiv preprint arXiv:2302.13861, 2023. 1

    2023 arXiv

  4. [12]

    Dp-merf: Differentially private mean embeddings with ran- domfeatures for practical privacy-preserving data genera- tion

    Frederik Harder, Kamil Adamczewski, and Mijung Park. Dp-merf: Differentially private mean embeddings with ran- domfeatures for practical privacy-preserving data genera- tion. In International conference on artificial intelligence and statistics, pages 1819–1827. PMLR, 2021. 6, 8

    2021

  5. [13]

    Adam: A method for stochastic opti- mization

    Diederik P Kingma. Adam: A method for stochastic opti- mization. arXiv preprint arXiv:1412.6980, 2014. 4

    2014 arXiv

  6. [14]

    Imagenet classification with deep convolutional neural net- works

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural net- works. NeurIPS, 25, 2012. 8

    2012

  7. [15]

    Dataset condensation with contrastive signals

    Saehyung Lee and Sung Ju Hwang. Dataset condensation with contrastive signals. In Proceedings of the 39th Inter- national Conference on Machine Learning , pages 12244– 12254. PMLR, 2022. 3

    2022

  8. [16]

    Dataset distillation via the wasserstein metric.arXiv preprint arXiv:2311.18531,

    Haoyang Liu, Yijiang Li, Tiancheng Xing, Vibhu Dalal, Luwei Li, Jingrui He, and Haohan Wang. Dataset distillation via the wasserstein metric.arXiv preprint arXiv:2311.18531,

    arXiv

  9. [17]

    Efficient dataset distillation using random feature approximation

    Noel Loo, Akash Vasudevan, Alexandre Bayen, and Brandon Malone. Efficient dataset distillation using random feature approximation. In Advances in Neural Information Process- ing Systems, pages 1–13, 2022. 3

    2022

  10. [18]

    Dataset distillation with convexified implicit gra- dients

    Noel Loo, Akash Vasudevan, Brandon Malone, and Alexan- dre Bayen. Dataset distillation with convexified implicit gra- dients. In Proceedings of the 40th International Conference on Machine Learning, pages 1–15. PMLR, 2023. 3

    2023

  11. [19]

    R ´enyi differential privacy

    Ilya Mironov. R ´enyi differential privacy. InIEEE CSF, pages 263–275. IEEE, 2017. 2

    2017

  12. [20]

    Dataset distillation with infinitely wide convolutional networks

    Timothy Nguyen, Roman Novak, Jaehoon Lee, and Lechao Xiao. Dataset distillation with infinitely wide convolutional networks. In Advances in Neural Information Processing Systems, pages 5184–5197, 2021. 3

    2021

  13. [21]

    Dataset meta-learning from kernel ridge-regression

    Timothy Nguyen, Roman Novak, Lechao Xiao, and Jaehoon Lee. Dataset meta-learning from kernel ridge-regression. In International Conference on Learning Representations ,

  14. [22]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022. 6, 7, 4

    2022

  15. [23]

    Data distillation: A survey, 2023

    Noveen Sachdeva and Julian McAuley. Data distillation: A survey, 2023. 2

    2023

  16. [24]

    DataDAM: Efficient dataset distillation with attention matching

    Ahmad Sajedi, Samir Khaki, Rui Wang, et al. DataDAM: Efficient dataset distillation with attention matching. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023. 3

    2023

  17. [25]

    Laion-5b: An open large-scale dataset for training next generation image-text models

    Christoph Schuhmann, Romain Beaumont, Richard Vencu, Cade Gordon, Ross Wightman, Mehdi Cherti, Theo Coombes, Aarush Katta, Clayton Mullis, Mitchell Worts- man, et al. Laion-5b: An open large-scale dataset for training next generation image-text models. Advances in Neural In- f...

    2022

  18. [26]

    Loss-curvature matching for dataset selection and condensation

    Seungjae Shin, Heesun Bae, Sangwoo Kim, and Taesup Moon. Loss-curvature matching for dataset selection and condensation. In International Conference on Artificial In- telligence and Statistics, pages 1–14, 2023. 3

    2023

  19. [27]

    Denois- ing diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denois- ing diffusion implicit models. In International Conference on Learning Representations. 4

  20. [28]

    Score-based generative modeling through stochastic differential equa- tions

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Ab- hishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equa- tions. In International Conference on Learning Represen- tations, 2021. 4

    2021

  21. [29]

    Differentially pri- vate kernel inducing points using features from scatternets (DP-KIP-scatternet) for privacy preserving data distillation

    Margarita Vinaroz and Mijung Park. Differentially pri- vate kernel inducing points using features from scatternets (DP-KIP-scatternet) for privacy preserving data distillation. Transactions on Machine Learning Research, 2024. 3, 6, 8

    2024

  22. [30]

    Cafe: Learning to condense dataset by align- ing features

    Kai Wang, Bo Zhao, Xiangyu Peng, Zheng Zhu, Shuo Yang, Shuo Wang, Guan Huang, Hakan Bilen, Xinchao Wang, and Yang You. Cafe: Learning to condense dataset by align- ing features. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 12196– ...

    2022

  23. [31]

    Dataset distillation

    Tongzhou Wang, Jun-Yan Zhu, Antonio Torralba, and Alexei A Efros. Dataset distillation. arXiv preprint arXiv:1811.10959, 2018. 1, 2

    2018 arXiv

  24. [32]

    Subsampled r ´enyi differential privacy and analytical moments accountant

    Yu-Xiang Wang, Borja Balle, and Shiva Prasad Ka- siviswanathan. Subsampled r ´enyi differential privacy and analytical moments accountant. In The 22nd International Conference on Artificial Intelligence and Statistics , pages 1226–1235. PMLR, 2019. 2

    2019

  25. [33]

    Fashion- mnist: a novel image dataset for benchmarking machine learning algorithms, 2017

    Han Xiao, Kashif Rasul, and Roland V ollgraf. Fashion- mnist: a novel image dataset for benchmarking machine learning algorithms, 2017. 6

    2017

  26. [34]

    Opacus: User-friendly differential privacy library in pytorch

    Ashkan Yousefpour, Igor Shilov, Alexandre Sablayrolles, Davide Testuggine, Karthik Prasad, Mani Malek, John Nguyen, Sayan Ghosh, Akash Bharadwaj, Jessica Zhao, et al. Opacus: User-friendly differential privacy library in pytorch. arXiv preprint arXiv:2109.12298, 2021. 4

    2021 arXiv

  27. [35]

    M3D: Dataset condensation by minimizing maximum mean discrepancy

    Hansong Zhang, Shikun Li, Bo Zhao, et al. M3D: Dataset condensation by minimizing maximum mean discrepancy. In Proceedings of the AAAI Conference on Artificial Intel- ligence, 2024. 3

    2024

  28. [36]

    Dataset condensation with dif- ferentiable siamese augmentation

    Bo Zhao and Hakan Bilen. Dataset condensation with dif- ferentiable siamese augmentation. In Proceedings of the 38th International Conference on Machine Learning , pages 12674–12685. PMLR, 2021. 2

    2021

  29. [37]

    Dataset condensation with gradient matching

    Bo Zhao, Konda Reddy Mopuri, and Hakan Bilen. Dataset condensation with gradient matching. In International Con- ference on Learning Representations, 2021. 2, 4

    2021

  30. [38]

    Improved distri- bution matching for dataset condensation

    Ganlong Zhao, Bo Zhao, and Hakan Bilen. Improved distri- bution matching for dataset condensation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4903–4912, 2023. 6, 8

    2023

  31. [39]

    Differentially private dataset condensation

    Tianhang Zheng and Baochun Li. Differentially private dataset condensation. 2022. 2, 3, 4, 6, 8

    2022

  32. [40]

    w/” and “w/o

    Yongchao Zhou, Jianfei Wang, Jian Chen, et al. Dataset dis- tillation using neural feature regression. In Advances in Neu- ral Information Processing Systems, pages 1–12, 2022. 3 Improving Noise Efficiency in Privacy-preserving Dataset Distillation Supplementary Material A. Th...

    2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.