mlsanityData to 14 Oct 2026

Private Multi-Winner Voting for Machine Learning

arXiv 2211.15410 ↗First posted 23 Nov 2022cs.LG cs.CRcited by 0

Not on the map. Why?

Abstract (arXiv)

Private multi-winner voting is the task of revealing k-hot binary vectors satisfying a bounded differential privacy (DP) guarantee. This task has been understudied in machine learning literature despite its prevalence in many domains such as healthcare. We propose three new DP multi-winner mechanisms: Binary, τ, and Powerset voting. Binary voting operates independently per label through composition. τ voting bounds votes optimally in their ℓ2 norm for tight data-independent guarantees. Powerset voting operates over the entire binary vector by viewing the possible outcomes as a power set. Our theoretical and empirical analysis shows that Binary voting can be a competitive mechanism on many tasks unless there are strong correlations between labels, in which case Powerset voting outperforms it. We use our mechanisms to enable privacy-preserving multi-label learning in the central setting by extending the canonical single-label technique: PATE. We find that our techniques outperform current state-of-the-art approaches on large, real-world healthcare data and standard multi-label benchmarks. We further enable multi-label confidential and private collaborative (CaPC) learning and show that model performance can be significantly improved in the multi-site setting.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 32 of its 71 references that are in the map. First posted on arXiv 23 Nov 2022.

Builds on

32 of 71 references.

References · 32

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2016Densely Connected Convolutional Networkscited by 6,720
  3. 2016Deep Learning with Differential Privacycited by 2,728
  4. 2019CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparisoncited by 1,423
  5. 2019MixMatch: A Holistic Approach to Semi-Supervised Learningcited by 1,391
  6. 2017Renyi Differential Privacycited by 650
  7. 2019MIMIC-CXR-JPG, a large publicly available database of labeled chest radiographscited by 597
  8. 2020Label-Only Membership Inference Attackscited by 388
  9. 2020Asymmetric Loss For Multi-Label Classificationcited by 381
  10. 2016Semi-supervised Knowledge Transfer for Deep Learning from Private Training Datacited by 348
  11. 2019PadChest: A large chest x-ray image dataset with multi-label annotated reportscited by 309
  12. 2013The Composition Theorem for Differential Privacycited by 297
20 more
  1. 2018Scalable Private Learning with PATEcited by 269
  2. 2021Adversary Instantiation: Lower Bounds for Differentially Private Machine Learningcited by 199
  3. 2020GS-WGAN: A Gradient-Sanitized Approach for Learning Differentially Private Generatorscited by 117
  4. 2021Hyperparameter Tuning with Renyi Differential Privacycited by 104
  5. 2020On the limits of cross-domain generalization in automated X-ray predictioncited by 62
  6. 2020Enabling Fast Differentially Private SGD via Just-in-Time Compilation and Vectorizationcited by 58
  7. 2020Interpreting Chest X-rays via CNNs that Exploit Hierarchical Disease Dependencies and Uncertainty Labelscited by 56
  8. 2020Chasing Your Long Tails: Differentially Private Prediction in Health Care Settingscited by 37
  9. 2021CaPC Learning: Confidential and Private Collaborative Learningcited by 33
  10. 2019Federated Heavy Hitters Discovery with Differential Privacycited by 32
  11. 2017Practical Locally Private Heavy Hitterscited by 29
  12. 2015LSHTC: A Benchmark for Large-Scale Text Classificationcited by 22
  13. 2016Differentially Private Random Decision Forests using Smooth Sensitivitycited by 16
  14. 2020A bounded-noise mechanism for differential privacycited by 15
  15. 2020Multi-label classification: do Hamming loss and subset accuracy really conflict with each other?cited by 14
  16. 2019What Do Multiwinner Voting Rules Do? An Experiment Over the Two-Dimensional Euclidean Domaincited by 10
  17. 2021PATE-AAE: Incorporating Adversarial Autoencoder into Private Aggregation of Teacher Ensembles for Spoken Command Classificationcited by 10
  18. 2020Extreme Multi-label Classification from Aggregated Labelscited by 5
  19. 2014Differentially- and non-differentially-private random decision treescited by 4
  20. 2019Data-Dependent Differentially Private Parameter Learning for Directed Graphical Modelscited by 3

Built on by

Not cited yet.