ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 2204.10536
19
3

Sharper Utility Bounds for Differentially Private Models

22 April 2022
Yilin Kang
Yong Liu
Jian Li
Weiping Wang
    FedML
ArXivPDFHTML
Abstract

In this paper, by introducing Generalized Bernstein condition, we propose the first O(pnϵ)\mathcal{O}\big(\frac{\sqrt{p}}{n\epsilon}\big)O(nϵp​​) high probability excess population risk bound for differentially private algorithms under the assumptions GGG-Lipschitz, LLL-smooth, and Polyak-{\L}ojasiewicz condition, based on gradient perturbation method. If we replace the properties GGG-Lipschitz and LLL-smooth by α\alphaα-H{\"o}lder smoothness (which can be used in non-smooth setting), the high probability bound comes to O(n−α1+2α)\mathcal{O}\big(n^{-\frac{\alpha}{1+2\alpha}}\big)O(n−1+2αα​) w.r.t nnn, which cannot achieve O(1/n)\mathcal{O}\left(1/n\right)O(1/n) when α∈(0,1]\alpha\in(0,1]α∈(0,1]. To solve this problem, we propose a variant of gradient perturbation method, \textbf{max{1,g}\{1,g\}{1,g}-Normalized Gradient Perturbation} (m-NGP). We further show that by normalization, the high probability excess population risk bound under assumptions α\alphaα-H{\"o}lder smooth and Polyak-{\L}ojasiewicz condition can achieve O(pnϵ)\mathcal{O}\big(\frac{\sqrt{p}}{n\epsilon}\big)O(nϵp​​), which is the first O(1/n)\mathcal{O}\left(1/n\right)O(1/n) high probability excess population risk bound w.r.t nnn for differentially private algorithms under non-smooth conditions. Moreover, we evaluate the performance of the new proposed algorithm m-NGP, the experimental results show that m-NGP improves the performance of the differentially private model over real datasets. It demonstrates that m-NGP improves the utility bound and the accuracy of the DP model on real datasets simultaneously.

View on arXiv
Comments on this paper