mlsanityData to 7 Oct 2026

Mean Estimation with User-level Privacy under Data Heterogeneity

arXiv 2307.15835 ↗First posted 28 Jul 2023cs.CR cs.DS cs.LGcited by 8

Not on the map. Why?

Abstract (arXiv)

A key challenge in many modern data analysis tasks is that user data are heterogeneous. Different users may possess vastly different numbers of data points. More importantly, it cannot be assumed that all users sample from the same underlying distribution. This is true, for example in language data, where different speech styles result in data heterogeneity. In this work we propose a simple model of heterogeneous user data that allows user data to differ in both distribution and quantity of data, and provide a method for estimating the population-level mean while preserving user-level differential privacy. We demonstrate asymptotic optimality of our estimator and also prove general lower bounds on the error achievable in the setting we introduce.

Key facts. Cited by 8 papers in the map (data through 7 Oct 2026). Builds on 18 of its 35 references that are in the map. First posted on arXiv 28 Jul 2023.

Builds on

18 of 35 references.

References · 18

  1. 2016Communication-Efficient Learning of Deep Networks from Decentralized Datacited by 7,592
  2. 2019Advances and Open Problems in Federated Learningcited by 3,021
  3. 2014RAPPOR: Randomized Aggregatable Privacy-Preserving Ordinal Responsecited by 262
  4. 2020Federated Learning of a Mixture of Global and Local Modelscited by 208
  5. 2018Protection Against Reconstruction and Its Applications in Private Federated Learningcited by 127
  6. 2016Minimax Optimal Procedures for Locally Private Estimationcited by 103
  7. 2017Finite Sample Differentially Private Confidence Intervalscited by 103
  8. 2017On the convergence properties of a K-step averaging stochastic gradient descent algorithm for nonconvex optimizationcited by 91
  9. 2021Learning with User-Level Privacycited by 72
  10. 2020Breaking the Communication-Privacy-Accuracy Trilemmacited by 68
  11. 2020Learning discrete distributions: user vs item-level privacycited by 41
  12. 2019Semi-Cyclic Stochastic Gradient Descentcited by 40
6 more
  1. 2015Simultaneous Private Learning of Multiple Conceptscited by 34
  2. 2020Differentially Private Simple Linear Regressioncited by 33
  3. 2018Hadamard Response: Estimating Distributions Privately, Efficiently, and with Little Communicationcited by 26
  4. 2019Communication Complexity in Locally Private Distribution Estimation and Heavy Hitterscited by 11
  5. 2022Instance-Optimal Differentially Private Estimationcited by 7
  6. 2022A Generative Framework for Personalized Learning and Estimation: Theory, Algorithms, and Privacycited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 2; 2024: 4; 2025: 1; 2026: 1.

8 papers.

  1. 2024Fine-Tuning Large Language Models with User-Level Differential Privacycited by 28
  2. 2023Better and Simpler Lower Bounds for Differentially Private Statistical Estimationcited by 16
  3. 2024A Huber Loss Minimization Approach to Mean Estimation under User-level Differential Privacycited by 11
  4. 2023Mean Estimation Under Heterogeneous Privacy Demandscited by 6
  5. 2024Privacy Preserving Semi-Decentralized Mean Estimation over Intermittently-Connected Networkscited by 2
  6. 2024Better Locally Private Sparse Estimation Given Multiple Samples Per Usercited by 1
  7. 2025Black-Box Privacy Attacks on Shared Representations in Multitask Learningcited by 0
  8. 2026Matrix Factorization for Practical Continual Mean Estimation Under User-Level Differential Privacycited by 0