Pith. sign in

REVIEW 7 major objections 5 minor 40 references

Aggregating Data for Optimal and Private Learning

T0 review · 7 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read This paper establishes that for MIR and LLP under squared loss, the bagging configuration maximizing downstream linear-regression utility is approximately the one minimizing k-means-like objectives over expected labels or bag-centroid…

desk verdict Useful results on bag construction for MIR/LLP under squared loss, but the headline optimality claim rests on an unstated equal-size assumption, and the private LLP bound is internally inconsistent. read the letter →

arxiv 2411.19045 v1 pith:5R3IQSES submitted 2024-11-28 cs.LG

Sushant Agarwal , Yukti Makhija , Rishi Saket , Aravindan Raghuveer This is my paper · ORCID
classification cs.LG
keywords multipleinstanceregressionlearningfromlabelproportionsbaggingk-meansclusteringdifferentialprivacylineargeneralizedmodelsutilitybounds
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 asks how a trusted aggregator should partition data into bags when learners only see aggregate labels—one random label per bag (multiple instance regression, MIR) or the bag mean (learning from label proportions, LLP)—so that a downstream linear-regression model is as accurate as possible. The authors prove that for each setting the bagging that maximizes utility is approximately the one minimizing a natural clustering objective: a k-means objective over the expected labels for instance-level MIR, the condition number of the bag-centroid covariance for bag-level LLP, and both for aggregate-level MIR. They further show that each bagging mechanism can be made label-differentially private by adding calibrated noise, with the extra error quantified in the utility bounds, and they extend the analysis to generalized linear models. The upshot is that a single, well-studied primitive—k-means clustering—can serve as the organizing principle for private data aggregation in regression.

What carries the argument

The argument runs through two linear-algebra objects that encode bagging: the attribution matrix $A$ for instance-MIR, which assigns each bag's disclosed label to every instance in the bag (with expectation $S$ having entries $1/|B_l|$), and the bagging matrix $S$ for bag-LLP and aggregate-MIR, whose rows average the feature vectors in each bag. The utility bounds are obtained by expanding $\hat{\theta} - \theta^*$ around the true model and bounding terms with the operator norm of $(X^TX)^{-1}X^T$, the expected attribution error, and the spectrum of $(SX)^T(SX)$. The load-bearing identity is Lemma 13: minimizing the k-means objective over expected labels is exactly equivalent to maximizing $\sum_l (\sum_{i \in B_l} \tilde{y}_i)^2 / |B_l|$, the bag-dependent term in Theorem 1. The LLP and aggregate bounds hinge on the condition number of the bag-centroid covariance, which random bagging controls via a matrix Chernoff tail bound.

What would settle it

Take four points with expected labels 0, 0, 10, 10, form two bags with minimum size 1, and compare the Theorem 1 utility-bound term. Equal bags of size 2 give $(0+0)^2/2 + (10+10)^2/2 = 200$, while unequal bags of sizes 1 and 3 give $0^2/1 + (0+10+10)^2/3 \approx 133.3$, so the unequal configuration beats the equal-size solution that the k-means reduction selects.

Watch

Extended reading notes

Core claim

Under squared loss with labels drawn as $y_i = x_i^T \theta^* + \gamma_i$, the paper establishes utility upper bounds on $\mathbb{E}[\|\hat{\theta} - \theta^*\|_2^2]$ for three bagged-regression estimators. For instance-MIR, the bound (Theorem 1) is minimized exactly by the k-means clustering of expected labels $\tilde{y}$, and since $\tilde{y}$ is a one-dimensional projection, the optimal bags are contiguous segments of the sorted labels. For bag-level LLP, the bound (Theorem 2) reduces to minimizing the ratio $\lambda_{\max}/\lambda_{\min}$ of the covariance matrix of bag centroids—a label-agnostic objective—and a random bagging algorithm provides a high-probability upper bound on this ratio. For aggregate-level MIR, the bound (Theorem 3) combines the label k-means term with the centroid-condition-number term. The paper then shows that these baggings remain effective under label differential privacy: private instance-MIR pays an extra error from private clustering plus subsampled noise on bag labels, while private bag-LLP pays only the label-noise error because its bagging is label-agnostic. Experiments on synthetic data confirm that instance k-means is a strong heuristic across all three settings.

Load-bearing premise

The exact reduction of instance-MIR to k-means requires every bag to have exactly the same size k, the constraint imposed in equation (3), whereas the problem only requires bags of size at least k; if the true optimum uses unequal bag sizes, the claimed equivalence is not established.

Editorial extensions

If this is right

  • If the theorems hold, a practitioner building a regression model from aggregate labels can treat 'choose bags' as 'run k-means on labels or features' and obtain a utility guarantee with no further search.
  • For bag-level LLP, the optimal bagging is label-agnostic, so the same bagging can be released without using label information, which matters when the aggregator may not know the labels.
  • Label differential privacy is achievable at a quantified cost: private instance-MIR pays an extra term from private clustering and subsampled label noise, while private bag-LLP pays only label noise.
  • The GLM extension implies that for instance-MIR over the exponential family, label k-means remains the right bagging target, and for aggregate-MIR with a monotone link, grouping instances with similar expected labels into contiguous sorted bags is optimal.
  • Since 1D k-means with equal-size bags has contiguous sorted segments, the practical bagging algorithm is simply sorting by labels and cutting into equal chunks.

Reading between the lines

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

  • The equal-size assumption is the one place the k-means reduction could silently fail: the paper's experiments use equal-size bags, so whether unequal bags ever improve the non-private utility bound remains an open empirical question that the stated theorems do not settle.
  • The private-clustering step in Theorem 4 is analyzed via label noise that actually satisfies the stronger local-DP notion; switching to a central-model private k-means could shrink the $n(1-1/k)\alpha^2$ term, so the bound is likely loose in the privacy regime.
  • The condition-number objective for bag-LLP suggests that for non-isotropic data the effective label-agnostic bagging should cluster whitened features $\Sigma^{-1/2}X$ rather than raw $X$; the paper's scaled instance k-means experiments support this, but the theory's guarantee there is only heuristic.
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

7 major / 5 minor

Summary. The paper studies how to partition a training set into bags for Multiple Instance Regression (MIR) and Learning from Label Proportions (LLP) so that a downstream linear regression model, trained only on aggregate bag labels, is as close as possible to the underlying model. For instance-level MIR, bag-level LLP, and aggregate-level MIR, it derives upper bounds on the expected squared parameter error and claims that optimizing these bounds reduces to natural objectives: k-means over labels, minimizing the condition number of the bag-centroid covariance, and a combination of both. It then presents label-differentially-private versions of these bagging mechanisms with quantified extra error, extends the analysis to GLMs, and reports experiments on synthetic data.

Significance. If the reductions were proven, the paper would provide a principled, parameter-free account of optimal bag construction in aggregate-label regression, with practical guidance (k-means on feature vectors) and explicit privacy-utility trade-offs. The paper's strengths include the clean decomposition of the instance-MIR error into a bag-dependent k-means-like term, the breadth of the experimental study, and the attempt to give formal label-DP guarantees. However, the central optimality claims rely on several unproven equal-size reductions and a number of coefficient-level proof errors; the random-bagging analysis for bag-LLP is invalid as written, and the GLM extension is not quantitatively supported. The paper is promising but not ready in its current form.

major comments (7)
  1. [Section 1.2.1, Eq. (3); Theorem 1] The reduction to the equal-size k-means program (3) is not justified. The problem definition in Section 1.2 only requires |B_l| ≥ k and n ≥ mk; when n > mk, no feasible solution to (3) exists, because the program insists on |B_l| = k for all m bags. Lemma 13 is an identity for a fixed bagging and does not characterize the maximizer of T(B) = Σ_l (Σ_{i∈B_l} y~_i)^2 / |B_l| under the stated constraint; allowing variable bag sizes can increase T(B). Theorem 1 and the claimed equivalence therefore hold only under an unstated assumption such as n = mk, and the conclusion that optimal bagging reduces to k-means is not established for the problem as defined.
  2. [Section 2.2.1, Lemma 11 and Eq. (13)] The random-bagging analysis for bag-LLP is invalid as written. Lemma 11 states λ_min((SX)^T SX) = (1/k^2) λ_min(Σ_l X_l^T X_l), but (SX)^T SX is the sum of outer products of bag centroids, which includes cross terms (Σ_{i∈B_l} x_i)(Σ_{i∈B_l} x_i)^T; it is not equal to the sum of per-instance outer products X_l^T X_l. Equation (13) and the subsequent Matrix Chernoff bound therefore control a different matrix from the one appearing in Theorem 2. In addition, Theorem 2 defines f(X) = g(X)g(X)^T as an m×m matrix, which is singular with λ_min = 0 whenever m > d (as in the experiments, where m = 5,000 and d = 32), making the stated condition-number bound vacuous; the d×d matrix (SX)^T(SX) used in the appendix is the appropriate object.
  3. [Appendix B.2, Proposition 10] The proof of Proposition 10 uses E[||A||_op^2] = 1, but for the attribution matrix defined by Eq. (8), all rows in the same bag are identical unit vectors, so ||A||_op^2 = max_l |B_l|, which equals k for equal-size bags. This changes the constant in Theorem 1 and, more importantly, the bag-label noise term in Theorem 4: the claimed extra error d α^2/k^2 in the full version would become d k times the per-label noise variance (or an n term under the alternative inequality), so the privacy-utility trade-off is not correctly quantified.
  4. [Appendix B.1 and B.5.1, Lemma 14] Lemma 14 computes the expected difference between k-means objectives for a fixed bagging B, but the private clustering mechanism in Theorem 4 chooses B as the optimal clustering of the noisy labels, so B and the label noise γ are dependent. Conditioning on the realized B changes the distribution of γ, and the equality E[k-means(B(y))] = E[k-means(B(y~))] + (n-m)σ^2 does not follow for the data-dependent clustering. The claimed additional privacy error n(1-1/k)α^2 from 'add N(0,α^2) then cluster optimally' is therefore unsupported.
  5. [Appendix B.4, Theorem 3] In the proof of the aggregate-MIR bound, the term E[||Aγ||_2^2] is evaluated as nσ^2. Since A is the m×n attribution matrix of Eq. (15), Aγ contains one selected label noise per bag, so the expectation is mσ^2. This makes the noise term in Theorem 3 and in the aggregate-MIR privacy analysis (B.5.3, 'additional error of nk α^2/k^2') incorrect by a factor of k; the stated aggregate-MIR privacy guarantee is not established.
  6. [Section 1.2.4 vs. Appendix B.5.2, Theorem 5] Theorem 5 as stated in the main text, E[||·||^2] = OPT(d/k α^2 + σ^2 m/k) with α^2 = 4R^2 log(1.25/δ)/ε^2, differs from the full version in Appendix B.5.2, which gives OPT((σ^2 + α^2/k) m/k) = OPT(σ^2 m/k + α^2 m/k^2). The two formulas have different dependence on d and m, and the theorem is written as an equality even though Theorem 2 provides only an upper bound. The correct expression and its inequality status need to be reconciled.
  7. [Appendix F.1.1, Theorems 22 and 23] The GLM analysis contains dimensionally inconsistent algebra. In Theorem 22, the derived optimization objective is written as min Σ_l Σ_{i∈B_l} (μ_i - μ_l)^2 - Σ_l |B_l| μ_l, mixing a quadratic term with a linear term and misdefining μ_l (it appears to be μ_i/|B_l|); the remark that maximizing the second term forces larger bags contradicts the equal-size constraint used elsewhere. Moreover, the bound E[||Ay||^2] ≤ m(||b'(Xθ*)||^2 + ||Db''(Xθ*)||_1) is not derived; under equal-size bags the operator-norm factor should be k, not m. The aggregate-GLM bound in Theorem 23 contains an unexplained factor n and several missing justifications. The GLM generalization is not quantitatively supported in its current form.
minor comments (5)
  1. [Section 1.2 and Section 3] The relationship among n, m, and k is not stated consistently: the text says n ≥ mk, but the theorems reduce to equal-size bags with n = mk, and the experiments never report m. Please state the exact assumptions, including whether n is divisible by k.
  2. [Section 2.2.1, Algorithm 1] Algorithm 1 requires n to be divisible by 2k and silently outputs m/2 bags rather than m bags; for general n the procedure is undefined, and the relation to the m-bag problem is unclear.
  3. [Section 3, Table 2] The private experiments do not specify which differentially private k-means mechanism is used for label k-means, nor how the privacy budget is split between clustering and bag-label noise; without this, the empirical privacy results cannot be reproduced.
  4. [Section 1.2.4, Theorem 4] Theorem 4's displayed bound has unbalanced parentheses in the definition of α^2, and the main-text and appendix versions of the bound differ in the coefficient of OPT; these need to be reconciled.
  5. [References and typos] The reference list contains a broken citation '[Ska, pri]' and typos such as 'intital' in Appendix B.5.2 and 'event-level MIR' in Appendix F.1.1; these should be corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the k-means and condition-number reductions are algebraic identities derived from the stated utility bounds, and the privacy bounds use OPT as a data-defined optimization value, not a fitted parameter.

full rationale

The derivation chain is self-contained and does not reduce to its inputs. Theorem 1's reduction to label k-means follows from Proposition 7 and Lemma 13, which algebraically rewrite the bag-dependent utility term as ||tilde y||^2 minus the k-means objective; minimizing the bound is therefore equivalent to optimizing that objective under the equal-size constraint stated in equation (3). This is an identity, not a fitted parameter relabeled as a prediction. Theorems 2 and 3 similarly express utility bounds in terms of the condition number of the bag-centroid covariance and the label k-means term, and the proofs derive those quantities from the loss definitions. The privacy theorems (4 and 5) add calibrated noise and express the resulting utility loss in terms of OPT, where OPT is the value of the optimization objective over the data; it is an input to the bound, not a constant fitted from the data. The self-citations in the paper (Brahmbhatt et al., Saket et al., Chauhan et al.) are related-work references and none carries a load-bearing premise in the main derivation. One caveat is that optimization (3) imposes |B_l| = k, while the problem statement only requires |B_l| >= k; when n > mk the equal-size program is infeasible. This is a soundness and completeness concern about the claimed reduction, but it is not circularity, because the paper's algebra does not define the optimal bagging in terms of the k-means solution by construction. The core derivations are parameter-free from stated assumptions, so the appropriate circularity score is 0.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

The central claims rest on standard linear-algebra and concentration bounds. The main uncharged assumptions are the equal-size bag constraint, the boundedness of labels and features required for the DP and concentration results (R and β), and the monotonicity assumption in the GLM aggregate case.

free parameters (2)
  • R (label norm bound for label-DP) = not specified in experiments
    The label-DP results require labels bounded by R in absolute value; the experiments generate Gaussian labels from Xθ* plus noise with d=32, which are unbounded, and no clipping or scaling is described.
  • β (feature norm bound) = not specified
    Lemma 12 and the Matrix Chernoff argument in Section 2.2.1 assume ||x||² ≤ β for every feature vector; the experiments draw X from N(0,I), which has no deterministic norm bound.
assumptions (6)
  • domain assumption Labels follow y_i = x_i^T θ* + γ_i with γ_i ~ N(0, σ²)
    Used throughout Section 1.2 and the utility proofs; the derivations mainly need zero mean and covariance σ²I rather than normality.
  • standard math The design matrix X has rank d and n > d
    Required for (X^T X)^{-1} to exist in the least-squares estimators.
  • domain assumption Bags have minimum size |B_l| ≥ k, and Theorems 2 and 3 additionally assume equal size |B_l| = k
    The equal-size constraint is imposed for the LLP and aggregate-MIR bounds and appears without justification in the Theorem 1 reduction (equation (3)).
  • domain assumption For the aggregate-MIR GLM result, b'(·) is monotonic
    Appendix F.2 uses monotonicity of b' to upper-bound ||S b'(Xθ) - b'(S X θ)|| by the within-bag range of expected labels.
  • standard math Matrix Chernoff bound (Tropp 2012)
    Used for the high-probability lower bound on λ_min of the bag-centroid covariance in the random bagging analysis.
  • standard math Privacy amplification by subsampling (Balle et al. 2018)
    Justifies the noise scale added to bag labels in Theorems 4 and 5.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Aggregating Data for Optimal and Private Learning." pith.science (2026). https://pith.science/paper/5R3IQSES

@misc{pith2026241119045,
  author       = {Pith},
  title        = {Pith review of: Aggregating Data for Optimal and Private Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5R3IQSES}},
  note         = {Machine review of arXiv:2411.19045}
}
abstract

Multiple Instance Regression (MIR) and Learning from Label Proportions (LLP) are learning frameworks arising in many applications, where the training data is partitioned into disjoint sets or bags, and only an aggregate label i.e., bag-label for each bag is available to the learner. In the case of MIR, the bag-label is the label of an undisclosed instance from the bag, while in LLP, the bag-label is the mean of the bag's labels. In this paper, we study for various loss functions in MIR and LLP, what is the optimal way to partition the dataset into bags such that the utility for downstream tasks like linear regression is maximized. We theoretically provide utility guarantees, and show that in each case, the optimal bagging strategy (approximately) reduces to finding an optimal clustering of the feature vectors or the labels with respect to natural objectives such as $k$-means. We also show that our bagging mechanisms can be made label-differentially private, incurring an additional utility error. We then generalize our results to the setting of Generalized Linear Models (GLMs). Finally, we experimentally validate our theoretical results.

Figures

Figures reproduced from arXiv: 2411.19045 by the authors.

Figure 1
Figure 1. Random bagging algorithm for bag-LLP The feature matrix for bag B ′ l sampled using Algorithm 1 can be represented by X ′ l for all l ∈ [m′ ]. 1 k 2 λmin m ∑ l=1 X T l Xl ! ≥ 1 k 2 λmin m ∑ l=1 X ′ l T X ′ l ! (13) Let µmin = λmin  ∑ m′ l=1 E h X ′ l TX ′ l i /k 2 . We expand X ′ l TX ′ l and find µmin: µmin = 1 k 2 λmin   m′ ∑ l=1 E   ∑ xi ,xj∈B ′ l xix T j     = 1 k 2 λmin   m′ ∑ l=1 E   ∑ xi∈B ′ l x… view at source ↗
Figure 2
Figure 2. Random bagging algorithm for Agg-MIR Let Bl denote a super-bag of size 2k for l ∈ [m′ ]. We arbitrarily sample k instances to create a bag B (1) l and the remaining instances form another bag B (2) l . We know Bl = B (1) l S B (2) l and B (1) l T B (2) l = φ. Also, |B (1) l | = |B (2) l | = k. Theorem 20. For super-bags B ′ l as defined in Algorithm 2 with arbitrary non-overlapping partitions B (1) l and B (2) l , w… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 33 canonical work pages

  1. [1]

    https://developer.apple.com/documentation/storekit/skadnetwork/

    Apple storekit ad network. https://developer.apple.com/documentation/storekit/skadnetwork/

    work page

  2. [2]

    https://developer.chrome.com/docs/privacy-sandbox/aggregation-service/

    Private aggregation api of chrome privacy sandbox. https://developer.chrome.com/docs/privacy-sandbox/aggregation-service/

    work page

  3. [3]

    Co-training for demographic classification using deep learning from label proportions

    Ehsan Mohammady Ardehaly and Aron Culotta. Co-training for demographic classification using deep learning from label proportions. In ICDM , pages 1017--1024, 2017

    work page 2017

  4. [4]

    Privacy amplification by subsampling: Tight analyses via couplings and divergences, 2018

    Borja Balle, Gilles Barthe, and Marco Gaboardi. Privacy amplification by subsampling: Tight analyses via couplings and divergences, 2018. URL https://arxiv.org/abs/1807.01647

    arXiv 2018

  5. [5]

    Local differential privacy: a tutorial, 2019

    Björn Bebensee. Local differential privacy: a tutorial, 2019. URL https://arxiv.org/abs/1907.11908

    arXiv 2019

  6. [6]

    PAC learning linear thresholds from label proportions

    Anand Paresh Brahmbhatt, Rishi Saket, and Aravindan Raghuveer. PAC learning linear thresholds from label proportions. In Thirty-seventh Conference on Neural Information Processing Systems, 2023. URL https://openreview.net/forum?id=5Gw9YkJkFF

    work page 2023

  7. [7]

    Easy learning from label proportions

    Robert Istvan Busa-Fekete, Heejin Choi, Travis Dick, Claudio Gentile, and Andres Munoz medina. Easy learning from label proportions. arXiv, 2023. URL https://arxiv.org/abs/2302.03115

    arXiv 2023

  8. [8]

    Sample complexity bounds for differentially private learning

    Kamalika Chaudhuri and Daniel Hsu. Sample complexity bounds for differentially private learning. In Proceedings of the 24th Annual Conference on Learning Theory, pages 155--186. JMLR Workshop and Conference Proceedings, 2011

    work page 2011

Show all 40 references
  1. [9]

    Generalization and learnability in multiple instance regression

    Kushal Chauhan, Rishi Saket, Lorne Applebaum, Ashwinkumar Badanidiyuru, Chandan Giri, and Aravindan Raghuveer. Generalization and learnability in multiple instance regression. In UAI , 2024

    2024

  2. [10]

    L. Chen, Z. Huang, and R. Ramakrishnan. Cost-based labeling of groups of mass spectra. In Proc. ACM SIGMOD International Conference on Management of Data, pages 167--178, 2004

    2004

  3. [11]

    Learning from aggregated data: Curated bags versus random bags

    Lin Chen, Thomas Fu, Amin Karbasi, and Vahab Mirrokni. Learning from aggregated data: Curated bags versus random bags. arXiv, 2023. URL https://arxiv.org/abs/2305.09557

    2023 arXiv

  4. [12]

    de Freitas and H

    N. de Freitas and H. K \" u ck. Learning about individuals from group statistics. In Proc. UAI , pages 332--339, 2005

    2005

  5. [13]

    L. M. Dery, B. Nachman, F. Rubbo, and A. Schwartzman. Weakly supervised classification in high energy physics. Journal of High Energy Physics, 2017 0 (5): 0 1--11, 2017

    2017

  6. [14]

    Differential privacy

    Cynthia Dwork. Differential privacy. In Automata, Languages and Programming: 33rd International Colloquium, ICALP 2006, Venice, Italy, July 10-14, 2006, Proceedings, Part II 33, pages 1--12. Springer, 2006

    2006

  7. [15]

    Label differential privacy via clustering

    Hossein Esfandiari, Vahab Mirrokni, Umar Syed, and Sergei Vassilvitskii. Label differential privacy via clustering. In International Conference on Artificial Intelligence and Statistics, pages 7055--7075. PMLR, 2022

    2022

  8. [16]

    Deep learning with label differential privacy

    Badih Ghazi, Noah Golowich, Ravi Kumar, Pasin Manurangsi, and Chiyuan Zhang. Deep learning with label differential privacy. Advances in neural information processing systems, 34: 0 27131--27145, 2021

    2021

  9. [17]

    Regression with label differential privacy

    Badih Ghazi, Pritish Kamath, Ravi Kumar, Ethan Leeman, Pasin Manurangsi, Avinash Varadarajan, and Chiyuan Zhang. Regression with label differential privacy. arXiv preprint arXiv:2212.06074, 2022

    2022 arXiv

  10. [18]

    Priorboost: An adaptive algorithm for learning from aggregate responses, 2024

    Adel Javanmard, Matthew Fahrbach, and Vahab Mirrokni. Priorboost: An adaptive algorithm for learning from aggregate responses, 2024. URL https://arxiv.org/abs/2402.04987

    2024 arXiv

  11. [19]

    Kotzias, M

    D. Kotzias, M. Denil, N. de Freitas, and P. Smyth. From group to individual labels using deep features. In Proc. SIGKDD, pages 597--606, 2015

    2015

  12. [20]

    J. Liu, B. Wang, Z. Qi, Y. Tian, and Y. Shi. Learning from label proportions with generative adversarial networks. In Proc. NeurIPS , pages 7167--7177, 2019

    2019

  13. [21]

    Differentially private k-means clustering with convergence guarantee

    Zhigang Lu and Hong Shen. Differentially private k-means clustering with convergence guarantee. IEEE Transactions on Dependable and Secure Computing, page 1–1, 2020. ISSN 2160-9209. doi:10.1109/tdsc.2020.3043369. URL http://dx.doi.org/10.1109/TDSC.2020.3043369

    2020

  14. [22]

    Challenges and approaches to privacy preserving post-click conversion prediction

    Conor O'Brien, Arvind Thiagarajan, Sourav Das, Rafael Barreto, Chetan Verma, Tim Hsu, James Neufield, and Jonathan J Hunt. Challenges and approaches to privacy preserving post-click conversion prediction. arXiv preprint arXiv:2201.12666, 2022

    2022 arXiv

  15. [23]

    Quadrianto, A

    N. Quadrianto, A. J. Smola, T. S. Caetano, and Q. V. Le. Estimating labels from label proportions. J. Mach. Learn. Res., 10: 0 2349--2374, 2009

    2009

  16. [24]

    Ray and D

    S. Ray and D. Page. Multiple instance regression. In Proc. ICML , pages 425--432, 2001

    2001

  17. [25]

    Supervised versus multiple instance learning: an empirical comparison

    Soumya Ray and Mark Craven. Supervised versus multiple instance learning: an empirical comparison. In Proc. ICML, page 697–704, 2005

    2005

  18. [26]

    S. Rueping. SVM classifier estimation from group probabilities. In Proc. ICML, pages 911--918, 2010

    2010

  19. [27]

    R. Saket. Learnability of linear thresholds from label proportions. In Proc. NeurIPS, 2021. URL https://openreview.net/forum?id=5BnaKeEwuYk

    2021

  20. [28]

    R. Saket. Algorithms and hardness for learning linear thresholds from label proportions. In Proc. NeurIPS, 2022. URL https://openreview.net/forum?id=4LZo68TuF-4

    2022

  21. [29]

    On combining bags to better learn from label proportions

    Rishi Saket, Aravindan Raghuveer, and Balaraman Ravindran. On combining bags to better learn from label proportions. In AISTATS , volume 151 of Proceedings of Machine Learning Research, pages 5913--5927. PMLR , 2022. URL https://proceedings.mlr.press/v151/saket22a.html

    2022

  22. [30]

    Scott and J

    C. Scott and J. Zhang. Learning from label proportions: A mutual contamination framework. In Proc. NeurIPS, 2020

    2020

  23. [31]

    Composition of differential privacy & privacy amplification by subsampling, 2022

    Thomas Steinke. Composition of differential privacy & privacy amplification by subsampling, 2022. URL https://arxiv.org/abs/2210.00597

    2022 arXiv

  24. [32]

    Differentially private k -means clustering, 2015

    Dong Su, Jianneng Cao, Ninghui Li, Elisa Bertino, and Hongxia Jin. Differentially private k -means clustering, 2015. URL https://arxiv.org/abs/1504.05998

    2015 arXiv

  25. [33]

    Fuzzy and possibilistic clustering for multiple instance linear regression

    Mohamed Trabelsi and Hichem Frigui. Fuzzy and possibilistic clustering for multiple instance linear regression. In 2018 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE) , pages 1--7, 2018

    2018

  26. [34]

    User-friendly tail bounds for sums of random matrices

    Joel A Tropp. User-friendly tail bounds for sums of random matrices. Foundations of computational mathematics, 12: 0 389--434, 2012

    2012

  27. [35]

    A survey on semi-supervised learning

    Jesper E Van Engelen and Holger H Hoos. A survey on semi-supervised learning. Machine learning, 109 0 (2): 0 373--440, 2020

    2020

  28. [36]

    K. L. Wagstaff, T. Lane, and A. Roper. Multiple-instance regression with structured data. In Workshops Proceedings of the 8th IEEE ICDM , pages 291--300, 2008

    2008

  29. [37]

    Z. Wang, V. Radosavljevic, B. Han, Z. Obradovic, and S. Vucetic. Aerosol Optical Depth Prediction from Satellite Observations by Multiple Instance Regression, pages 165--176. 2008

    2008

  30. [38]

    Z. Wang, L. Lan, and S. Vucetic. Mixture model for multiple instance regression and applications in remote sensing. IEEE Transactions on Geoscience and Remote Sensing , 50 0 (6): 0 2226--2237, 2012

    2012

  31. [39]

    F. X. Yu, D. Liu, S. Kumar, T. Jebara, and S. F. Chang. SVM for learning with label proportions. In Proc. ICML, volume 28, pages 504--512, 2013

    2013

  32. [40]

    F. X. Yu, K. Choromanski, S. Kumar, T. Jebara, and S. F. Chang. On learning from label proportions. CoRR, abs/1402.5902, 2014. URL http://arxiv.org/abs/1402.5902

    2014 arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.