mlsanityData to 14 Oct 2026

Differentially Private Sharpness-Aware Training

arXiv 2306.05651 ↗First posted 9 Jun 2023cs.LG cs.AI cs.CRcited by 14

Not on the map. Why?

Abstract (arXiv)

Training deep learning models with differential privacy (DP) results in a degradation of performance. The training dynamics of models with DP show a significant difference from standard training, whereas understanding the geometric properties of private learning remains largely unexplored. In this paper, we investigate sharpness, a key factor in achieving better generalization, in private learning. We show that flat minima can help reduce the negative effects of per-example gradient clipping and the addition of Gaussian noise. We then verify the effectiveness of Sharpness-Aware Minimization (SAM) for seeking flat minima in private learning. However, we also discover that SAM is detrimental to the privacy budget and computational time due to its two-step optimization. Thus, we propose a new sharpness-aware training method that mitigates the privacy-optimization trade-off. Our experimental results demonstrate that the proposed method improves the performance of deep learning models with DP from both scratch and fine-tuning. Code is available at https://github.com/jinseongP/DPSAT.

Key facts. Cited by 14 papers in the map (data through 14 Oct 2026). Builds on 41 of its 55 references that are in the map. First posted on arXiv 9 Jun 2023.

Builds on

41 of 55 references.

References · 41

  1. 2020Training data-efficient image transformers & distillation through attentioncited by 4,586
  2. 2016Deep Learning with Differential Privacycited by 2,728
  3. 2020Sharpness-Aware Minimization for Efficiently Improving Generalizationcited by 1,417
  4. 2016On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minimacited by 1,352
  5. 2018Averaging Weights Leads to Wider Optima and Better Generalizationcited by 1,209
  6. 2017Visualizing the Loss Landscape of Neural Netscited by 1,130
  7. 2019Deep Leakage from Gradientscited by 1,021
  8. 2021CrossViT: Cross-Attention Multi-Scale Vision Transformer for Image Classificationcited by 806
  9. 2017Sharp Minima Can Generalize For Deep Netscited by 466
  10. 2021Gradient Descent on Neural Networks Typically Occurs at the Edge of Stabilitycited by 427
  11. 2021Large Language Models Can Be Strong Differentially Private Learnerscited by 412
  12. 2021Differentially Private Fine-tuning of Language Modelscited by 400
29 more
  1. 2021Opacus: User-Friendly Differential Privacy Library in PyTorchcited by 383
  2. 2019Differential Privacy Has Disparate Impact on Model Accuracycited by 316
  3. 2017Empirical Analysis of the Hessian of Over-Parametrized Neural Networkscited by 290
  4. 2021When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentationscited by 260
  5. 2020Differentially Private Learning Needs Better Features (or Much More Data)cited by 235
  6. 2019Differentially Private Learning with Adaptive Clippingcited by 220
  7. 2022Surrogate Gap Minimization Improves Sharpness-Aware Trainingcited by 172
  8. 2020Torchattacks: A PyTorch Repository for Adversarial Attackscited by 158
  9. 2020Implicit Gradient Regularizationcited by 146
  10. 2022Towards Understanding Sharpness-Aware Minimizationcited by 146
  11. 2022Generalized Federated Learning via Sharpness Aware Minimizationcited by 144
  12. 2021Efficient Sharpness-aware Minimization for Improved Training of Neural Networkscited by 137
  13. 2020Tempered Sigmoid Activations for Deep Learning with Differential Privacycited by 122
  14. 2022Penalizing Gradient Norm for Efficiently Improving Generalization in Deep Learningcited by 122
  15. 2021Sharpness-Aware Minimization Improves Language Model Generalizationcited by 108
  16. 2022Sharpness-Aware Training for Freecited by 95
  17. 2021Large Scale Private Learning via Low-rank Reparametrizationcited by 88
  18. 2021Do Not Let Privacy Overbill Utility: Gradient Embedding Perturbation for Private Learningcited by 86
  19. 2022When Do Flat Minima Optimizers Work?cited by 78
  20. 2019Measurements of Three-Level Hierarchical Structure in the Outliers in the Spectrum of Deepnet Hessianscited by 68
  21. 2023Gradient Norm Aware Minimization Seeks First-Order Flatness and Improves Generalizationcited by 57
  22. 2022Scalable and Efficient Training of Large Convolutional Neural Networks with Differential Privacycited by 48
  23. 2020Adaptive Inertia: Disentangling the Effects of Adaptive Learning Rate and Momentumcited by 40
  24. 2021On the Convergence and Calibration of Deep Learning with Differential Privacycited by 36
  25. 2021DPNAS: Neural Architecture Search for Deep Learning with Differential Privacycited by 20
  26. 2023Stability Analysis of Sharpness-Aware Minimizationcited by 14
  27. 2022A New Linear Scaling Rule for Private Adaptive Hyperparameter Optimizationcited by 11
  28. 2021DPlis: Boosting Utility of Differentially Private Deep Learning via Randomized Smoothingcited by 10
  29. 2023Exploring the Effect of Multi-step Ascent in Sharpness-Aware Minimizationcited by 9

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 1; 2024: 6; 2025: 2; 2026: 5.

14 papers.

Sort:

  1. 2024DiSK: Differentially Private Optimizer with Simplified Kalman Filter for Noise Reductioncited by 11
  2. 2024Beyond Local Sharpness: Communication-Efficient Global Sharpness-aware Minimization for Federated Learningcited by 8
  3. 2024DPAdapter: Improving Differentially Private Deep Learning through Noise Tolerance Pre-trainingcited by 7
  4. 2024Delving into Differentially Private Transformercited by 5
  5. 2023Bias-Aware Minimisation: Understanding and Mitigating Estimator Bias in Private SGDcited by 2
  6. 2024Federated Learning for Heterogeneous Electronic Health Record Systems with Cost Effective Participant Selectioncited by 1
  7. 2025Multi-Class Support Vector Machine with Differential Privacycited by 1
  8. 2025Optimizer Dynamics at the Edge of Stability with Differential Privacycited by 1
  9. 2026When Gradient Clipping Becomes a Control Mechanism for Differential Privacy in Deep Learningcited by 1
  10. 2024DP2-FedSAM: Enhancing Differentially Private Federated Learning Through Personalized Sharpness-Aware Minimizationcited by 0
  11. 2026Deep Learning under Fractional-Order Differential Privacycited by 0
  12. 2026Private and Stable Test-Time Adaptation with Differential Privacycited by 0
2 more
  1. 2026When Privacy Hurts Mergeability: Geometry-Aware Model Merging under Differential Privacycited by 0
  2. 2026Privacy Preserving Gossip Learningcited by 0