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. 2210.17070
34
5

Private optimization in the interpolation regime: faster rates and hardness results

31 October 2022
Hilal Asi
Karan N. Chadha
Gary Cheng
John C. Duchi
ArXivPDFHTML
Abstract

In non-private stochastic convex optimization, stochastic gradient methods converge much faster on interpolation problems -- problems where there exists a solution that simultaneously minimizes all of the sample losses -- than on non-interpolating ones; we show that generally similar improvements are impossible in the private setting. However, when the functions exhibit quadratic growth around the optimum, we show (near) exponential improvements in the private sample complexity. In particular, we propose an adaptive algorithm that improves the sample complexity to achieve expected error α\alphaα from dεα\frac{d}{\varepsilon \sqrt{\alpha}}εα​d​ to 1αρ+dεlog⁡(1α)\frac{1}{\alpha^\rho} + \frac{d}{\varepsilon} \log\left(\frac{1}{\alpha}\right)αρ1​+εd​log(α1​) for any fixed ρ>0\rho >0ρ>0, while retaining the standard minimax-optimal sample complexity for non-interpolation problems. We prove a lower bound that shows the dimension-dependent term is tight. Furthermore, we provide a superefficiency result which demonstrates the necessity of the polynomial term for adaptive algorithms: any algorithm that has a polylogarithmic sample complexity for interpolation problems cannot achieve the minimax-optimal rates for the family of non-interpolation problems.

View on arXiv
Comments on this paper