Research map / Federated learning and privacy / Differential privacy
Eliminating Solution Bias in Differentially Private Optimization
DONGRUN LI, YUN ZENG, Zibo Wei, Jiacheng Wei, Qinnan Zhang, Faguo Wu, Xiao Zhang, Zhiming Zheng
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use DP-C4 instead of DiceSGD | DP optimizer · CIFAR-10 (eps=5) | test accuracy (%): 60.24 -> 64.50 | Improved |
| use DP-C4 instead of DiceSGD | DP optimizer · Mushroom (SVM, logistic regression, eps=1) | test accuracy (%): 90.65 -> 96.98 | Improved |
| use DP-C4 instead of DiceSGD | DP optimizer · GLUE SST-2 (eps=2) | test accuracy (%): 78.71 -> 82.24 | Improved |
| use DP-C4 instead of DiceSGD | DP optimizer · IMDb (eps=2) | test accuracy (%): 78.19 -> 81.23 | Improved |
| use DP-C4 instead of DiceSGD | DP optimizer · CIFAR-100 (eps=5) | test accuracy (%): 40.73 -> 43.12 | Improved |
| use DP-C4 instead of DiceSGD | DP optimizer · Mnist (eps=5) | test accuracy (%): 97.02 -> 97.16 | No clear effect |
| use DP-C4 instead of DP-SGD | DP optimizer · CIFAR-10 (eps=5) | test accuracy (%): 53.05 -> 64.50 | Improved |
| use DP-SVRG variance reduction instead of DP-SGD | DP optimizer · Mushroom (SVM, logistic regression, eps=1) | test accuracy (%): 87.48 -> 77.13 | Hurt |
Closest papers by meaning
- Differentially Private SGD Without Clipping Bias: An Error-Feedback Approach (ICLR 2024 · similarity 0.92)
- Towards hyperparameter-free optimization with differential privacy (ICLR 2025 · similarity 0.91)
- DOPPLER: Differentially Private Optimizers with Low-pass Filter for Privacy Noise Reduction (NeurIPS 2024 · similarity 0.90)
- Online Sensitivity Optimization in Differentially Private Learning (AAAI 2024 · similarity 0.90)
- Differentially Private Stochastic Gradient Descent with Fixed-Size Minibatches: Tighter RDP Guarantees with or without Replacement (NeurIPS 2024 · similarity 0.90)
- Shifted Interpolation for Differential Privacy (ICML 2024 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
