DPSUR: Accelerating Differentially Private Stochastic Gradient Descent Using Selective Update and Release
Summary: DPSUR accelerates DP-SGD by validating noisy gradients and selectively applying only updates that improve convergence. Clipping-based randomization and thresholded selection address privacy, yielding faster convergence and higher utility across vision and text benchmarks. (summarized by gpt-5.6-luna on Jul 24 2026)
Incoming Non-self Citations Over Time
Authors
- 1. Jie Fu (East China Normal University; Hong Kong Polytechnic University)
- 2. Qingqing Ye (Hong Kong Polytechnic University)
- 3. Haibo Hu (Hong Kong Polytechnic University)
- 4. Zhili Chen (East China Normal University)
- 5. Lulu Wang (East China Normal University)
- 6. Kuncan Wang (East China Normal University)
- 7. Xun Ran (Hong Kong Polytechnic University)
BibTeX Citation
@article{fu_vldb24,
title = {{DPSUR: Accelerating Differentially Private Stochastic Gradient Descent Using Selective Update and Release}},
author = {Fu, Jie and Ye, Qingqing and Hu, Haibo and Chen, Zhili and Wang, Lulu and Wang, Kuncan and Ran, Xun},
journal = {PVLDB},
series = {{VLDB} '24},
volume = {17},
number = {6},
pages = {1200--1213},
doi = {10.14778/3648160.3648164},
url = {https://doi.org/10.14778/3648160.3648164},
year = {2024}
}Incoming Citations (Sorted by Pagerank)
Showing 2 of 2 citing papers.
| Rank | Citing Paper | Year | Venue | Pagerank |
|---|---|---|---|---|
| 5,746 | Federated Heavy Hitter Analytics with Local Differential Privacy | 2025 | SIGMOD | 6.1017514e-05 |
| 8,908 | Privacy and Accuracy-Aware AI/ML Model Deduplication | 2025 | SIGMOD | 5.3483178e-05 |
Previous
Page 1 / 1
Next
Outgoing Citations (Sorted by Pagerank)
Showing 1 of 1 cited papers.
Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.
| Rank | Cited Paper | Year | Venue | Pagerank |
|---|---|---|---|---|
| 11,444 | Trajectory Data Collection with Local Differential Privacy | 2023 | VLDB | 5.093636e-05 |
Previous
Page 1 / 1
Next
Semantically Similar Papers
| # | Overall Rank | Paper | Year | Venue |
|---|---|---|---|---|
| 1 | 13,406 | High Dimensional Differentially Private Stochastic Optimization with Heavy-tailed Data | 2022 | PODS |
| 2 | 10,484 | Privacy Loss of Noise Perturbation via Concentration Analysis of A Product Measure | 2026 | SIGMOD |
| 3 | 8,311 | DP-Sync: Hiding Update Patterns in Secure Outsourced Databases with Differential Privacy | 2021 | SIGMOD |
| 4 | 6,406 | Skellam Mixture Mechanism: a Novel Approach to Federated Learning with Differential Privacy | 2022 | VLDB |
| 5 | 7,841 | Quantifying identifiability to choose and audit epsilon in differentially private deep learning | 2021 | VLDB |
| 6 | 10,335 | A General Framework for Per-record Differential Privacy | 2026 | SIGMOD |
| 7 | 6,366 | Practical Differentially Private and Byzantine-resilient Federated Learning | 2023 | SIGMOD |
| 8 | 7,517 | Private Incremental Regression | 2017 | PODS |
| 9 | 4,185 | Bolt-on Differential Privacy for Scalable Stochastic Gradient Descent-based Analytics | 2017 | SIGMOD |
| 10 | 8,908 | Privacy and Accuracy-Aware AI/ML Model Deduplication | 2025 | SIGMOD |