Differentially Private Adaptive Optimization with Delayed Preconditioners

• URL: http://arxiv.org/abs/2212.00309v2
• Date: Wed, 7 Jun 2023 20:22:57 GMT
• Title: Differentially Private Adaptive Optimization with Delayed Preconditioners
• Authors: Tian Li, Manzil Zaheer, Ken Ziyu Liu, Sashank J. Reddi, H. Brendan McMahan, Virginia Smith
• Abstract summary: We explore techniques to estimate adapt geometry in training without auxiliary data. Motivated by the observation that adaptive methods can tolerate stale preconditioners, we propose differentially adaptively private training. Empirically, we explore DP2, demonstrating that it can improve convergence speed by as much as 4x relative to non-adaptive baselines.
• Score: 44.190582378775694
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Privacy noise may negate the benefits of using adaptive optimizers in differentially private model training. Prior works typically address this issue by using auxiliary information (e.g., public data) to boost the effectiveness of adaptive optimization. In this work, we explore techniques to estimate and efficiently adapt to gradient geometry in private adaptive optimization without auxiliary data. Motivated by the observation that adaptive methods can tolerate stale preconditioners, we propose differentially private adaptive training with delayed preconditioners (DP^2), a simple method that constructs delayed but less noisy preconditioners to better realize the benefits of adaptivity. Theoretically, we provide convergence guarantees for our method for both convex and non-convex problems, and analyze trade-offs between delay and privacy noise reduction. Empirically, we explore DP^2 across several real-world datasets, demonstrating that it can improve convergence speed by as much as 4x relative to non-adaptive baselines and match the performance of state-of-the-art optimization methods that require auxiliary data.

Related papers

• Adaptive Preference Scaling for Reinforcement Learning with Human Feedback [103.36048042664768]
  Reinforcement learning from human feedback (RLHF) is a prevalent approach to align AI systems with human values. We propose a novel adaptive preference loss, underpinned by distributionally robust optimization (DRO) Our method is versatile and can be readily adapted to various preference optimization frameworks.
  arXiv  Detail & Related papers  (2024-06-04T20:33:22Z)
• Adaptive Bayesian Optimization for High-Precision Motion Systems [2.073673208115137]
  We propose a real-time purely data-driven, model-free approach for adaptive control, by online tuning low-level controller parameters. We base our algorithm on GoOSE, an algorithm for safe and sample-efficient Bayesian optimization. We evaluate the algorithm's performance on a real precision-motion system utilized in semiconductor industry applications.
  arXiv  Detail & Related papers  (2024-04-22T21:58:23Z)
• Online Sensitivity Optimization in Differentially Private Learning [8.12606646175019]
  We present a novel approach to dynamically optimize the clipping threshold. We treat this threshold as an additional learnable parameter, establishing a clean relationship between the threshold and the cost function. Our method is thoroughly assessed against alternative fixed and adaptive strategies across diverse datasets, tasks, model dimensions, and privacy levels.
  arXiv  Detail & Related papers  (2023-10-02T00:30:49Z)
• Adaptive Gradient Methods with Local Guarantees [48.980206926987606]
  We propose an adaptive gradient method that has provable adaptive regret guarantees vs. the best local preconditioner. We demonstrate the robustness of our method in automatically choosing the optimal learning rate schedule for popular benchmarking tasks in vision and language domains.
  arXiv  Detail & Related papers  (2022-03-02T20:45:14Z)
• Private Adaptive Optimization with Side Information [48.91141546624768]
  AdaDPS is a general framework that uses non-sensitive side information to precondition the gradients. We show AdaDPS reduces the amount of noise needed to achieve similar privacy guarantees. Our results show that AdaDPS improves accuracy by 7.7% (absolute) on average.
  arXiv  Detail & Related papers  (2022-02-12T03:02:06Z)
• Adaptive Differentially Private Empirical Risk Minimization [95.04948014513226]
  We propose an adaptive (stochastic) gradient perturbation method for differentially private empirical risk minimization. We prove that the ADP method considerably improves the utility guarantee compared to the standard differentially private method in which vanilla random noise is added.
  arXiv  Detail & Related papers  (2021-10-14T15:02:20Z)
• Adaptive Gradient Methods for Constrained Convex Optimization and Variational Inequalities [32.51470158863247]
  Main algorithms AdaACSA and AdaAGD+ are accelerated methods for constrained convex optimization. We complement them with a simpler algorithm AdaGrad+ which enjoys the same features, and achieves the standard non-accelerated convergence rate.
  arXiv  Detail & Related papers  (2020-07-17T09:10:21Z)
• Bilevel Optimization for Differentially Private Optimization in Energy Systems [53.806512366696275]
  This paper studies how to apply differential privacy to constrained optimization problems whose inputs are sensitive. The paper shows that, under a natural assumption, a bilevel model can be solved efficiently for large-scale nonlinear optimization problems.
  arXiv  Detail & Related papers  (2020-01-26T20:15:28Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.