Research map / Federated learning and privacy / Differential privacy
A(DP)^2SGD: Asynchronous Decentralized Parallel Stochastic Gradient Descent with Differential Privacy
Jie Jeff Xu, Wei Zhang, Wang Fei
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 |
|---|---|---|---|
| replace synchronous SGD with asynchronous decentralized A(DP)^2SGD under small noise | DP training · CIFAR-10 (ResNet-18) | test accuracy (%): 91.67 -> 91.64 | No clear effect |
| replace SYNC with A(DP)^2SGD under medium noise | DP training · CIFAR-10 (EfficientNet-B0) | test accuracy (%): 88.13 -> 87.01 | Hurt |
| replace SYNC with A(DP)^2SGD under large noise | DP training · CIFAR-10 (ResNet-18) | test accuracy (%): 83.40 -> 81.01 | Hurt |
| replace SYNC with A(DP)^2SGD under medium noise | DP training · SWB300 (LSTM) | held-out loss: 1.752 -> 1.732 | Improved |
| use ADPSGD instead of SYNC without noise | training · CIFAR-10 (ResNet-18) | test accuracy (%): 94.01 -> 94.97 | Improved |
| use ADPSGD instead of SYNC with a random learner 2X slower | training speed · CIFAR-10 | convergence vs runtime | Improved |
| use ADPSGD instead of SYNC with one learner 10X slower | training speed · CIFAR-10 | convergence vs runtime | Improved |
| use ADPSGD instead of SYNC with 2X batch size and 2X learning rate | training stability · CIFAR-10 (EfficientNet-B0) and SWB300 | convergence | Improved |
| A(DP)^2SGD at large noise for SENet-18 | SYNC · CIFAR-10 (SENet-18) | test accuracy (%): 10.00 | Hurt |
Closest papers by meaning
- Privacy-Preserving Federated Convex Optimization: Balancing Partial-Participation and Efficiency via Noise Cancellation (ICML 2025 · similarity 0.88)
- Mitigating the Privacy–Utility Trade-off in Decentralized Federated Learning via f-Differential Privacy (NeurIPS 2025 · similarity 0.88)
- The Privacy Power of Correlated Noise in Decentralized Learning (ICML 2024 · similarity 0.88)
- Private and Federated Stochastic Convex Optimization: Efficient Strategies for Centralized Systems (ICML 2024 · similarity 0.87)
- Differentially Private Stochastic Gradient Descent with Fixed-Size Minibatches: Tighter RDP Guarantees with or without Replacement (NeurIPS 2024 · similarity 0.87)
- Revisiting Differentially Private Algorithms for Decentralized Online Learning (ICML 2025 · similarity 0.87)
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.
