mlsanityData to 14 Oct 2026

Memory-Efficient Differentially Private Training with Gradient Random Projection

arXiv 2506.15588 ↗First posted 18 Jun 2025cs.LGcited by 2

Not on the map. Why?

Abstract (arXiv)

Differential privacy (DP) protects sensitive data during neural network training, but standard methods like DP-Adam suffer from high memory overhead due to per-sample gradient clipping, limiting scalability. We introduce DP-GRAPE (Gradient RAndom ProjEction), a DP training method that significantly reduces memory usage while maintaining utility on par with first-order DP approaches. DP-GRAPE is motivated by our finding that privatization flattens the gradient singular value spectrum, making SVD-based projections (as in GaLore (Zhao et al., 2024)) unnecessary. Consequently, DP-GRAPE employs three key components: (1) random Gaussian matrices replace SVD-based subspaces, (2) gradients are privatized after projection, and (3) projection is applied during backpropagation. These contributions eliminate the need for costly SVD computations, enable substantial memory savings, and lead to improved utility. Despite operating in lower-dimensional subspaces, our theoretical analysis shows that DP-GRAPE achieves a privacy-utility tradeoff comparable to DP-SGD. Our extensive empirical experiments show that DP-GRAPE can significantly reduce the memory footprint of DP training without sacrificing accuracy or training time. In particular, DP-GRAPE reduces memory usage by over 63% when pre-training Vision Transformers and over 70% when fine-tuning RoBERTa-Large as compared to DP-Adam, while achieving similar performance. We further demonstrate that DP-GRAPE scales to fine-tuning large models such as OPT with up to 6.7 billion parameters, a scale at which DP-Adam fails due to memory constraints. Our code is available at https://github.com/alexmul1114/DP_GRAPE.

Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Builds on 31 of its 42 references that are in the map. First posted on arXiv 18 Jun 2025.

Builds on

31 of 42 references.

References · 31

  1. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  2. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,380
  3. 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,372
  4. 2022OPT: Open Pre-trained Transformer Language Modelscited by 3,512
  5. 2016Deep Learning with Differential Privacycited by 2,728
  6. 2020Extracting Training Data from Large Language Modelscited by 2,189
  7. 2024GaLore: Memory-Efficient LLM Training by Gradient Low-Rank Projectioncited by 463
  8. 2021Large Language Models Can Be Strong Differentially Private Learnerscited by 412
  9. 2021Differentially Private Fine-tuning of Language Modelscited by 400
  10. 2023Fine-Tuning Language Models with Just Forward Passescited by 345
  11. 2019On Gradient Descent Ascent for Nonconvex-Concave Minimax Problemscited by 270
  12. 2022Unlocking High-Accuracy Differentially Private Image Classification through Scalecited by 224
19 more
  1. 2019Private Stochastic Convex Optimization with Optimal Ratescited by 186
  2. 2018Differentially Private Empirical Risk Minimization Revisited: Faster and More Generalcited by 128
  3. 2021Vision Transformer for Small-Size Datasetscited by 122
  4. 2024Flora: Low-Rank Adapters Are Secretly Gradient Compressorscited by 108
  5. 2020Bypassing the Ambient Dimension: Private SGD with Gradient Subspace Identificationcited by 90
  6. 2021Do Not Let Privacy Overbill Utility: Gradient Embedding Perturbation for Private Learningcited by 86
  7. 2022Differentially Private Optimization on Large Model at Small Costcited by 65
  8. 2020Scaling up Differentially Private Deep Learning with Fast Per-Example Gradient Clippingcited by 47
  9. 2022Memorization in NLP Fine-tuning Methodscited by 45
  10. 2024Private Fine-tuning of Large Language Models with Zeroth-order Optimizationcited by 41
  11. 2024Subspace Optimization for Large Language Models with Convergence Guaranteescited by 34
  12. 2022Faster Rates of Convergence to Stationary Points in Differentially Private Optimizationcited by 32
  13. 2021Fast and Memory Efficient Differentially Private-SGD via JL Projectionscited by 29
  14. 2023DPZero: Private Fine-Tuning of Language Models without Backpropagationcited by 26
  15. 2023Choosing Public Datasets for Private Machine Learning via Gradient Subspace Distancecited by 16
  16. 2024Addax: Utilizing Zeroth-Order Gradients to Improve Memory Efficiency and Performance of SGD for Fine-Tuning Language Modelscited by 11
  17. 2024How to Make the Gradients Small Privately: Improved Rates for Differentially Private Non-Convex Optimizationcited by 10
  18. 2023DP-Mix: Mixup-based Data Augmentation for Differentially Private Learningcited by 9
  19. 2024Pre-training Differentially Private Models with Limited Public Datacited by 8

Built on by

Cited by year

Citing papers per year: 2025: 2.

2 papers.

  1. 2025Forget What's Sensitive, Remember What Matters: Token-Level Differential Privacy in Memory Sculpting for Continual Learningcited by 1
  2. 2025Geometrically Principled Randomized Optimization for Efficient LLM Trainingcited by 1