Research map / Federated learning and privacy / Differential privacy
Make Landscape Flatter in Differentially Private Federated Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Add SAM local optimization to DP-FedAvg (ablation) | DP-FedAvg · EMNIST, Dirichlet 0.6 | validation accuracy (%): 83.20±0.01 -> 84.32±0.19 | Helped |
| Add SAM local optimization to Fed-SMP-topk (ablation) | Fed-SMP-topk · EMNIST, Dirichlet 0.6 | validation accuracy (%): 83.41±0.91 -> 84.80±0.60 | Helped |
| Add top-k sparsification to DP-FedSAM (ablation) | DP-FedSAM · CIFAR-10, IID | validation accuracy (%): 55.58±0.50 -> 58.82±0.51 | Helped |
| Add top-k sparsification to DP-FedSAM (ablation) | DP-FedSAM · CIFAR-10, Dirichlet 0.3 | validation accuracy (%): 53.92±0.55 -> 54.85±0.86 | Helped |
| Use top-k sparsification with DP-FedSAM rather than plain DP-FedSAM (ablation) | DP-FedSAM · CIFAR-10, Dirichlet 0.6 | test accuracy (%): 53.18 ± 0.40 -> 56.36 ± 0.36 | Helped |
| Increase the privacy budget ε | DP-FedSAM-topk · CIFAR-10, Dirichlet 0.6 | test accuracy (%): 38.96 ± 0.61 -> 49.17 ± 0.15 -> 53.64 ± 0.12 -> 56.36 ± 0.36 | Helped |
| Use a perturbation radius of ρ=0.5 rather than other tested radii (ablation) | DP-FedSAM · EMNIST, Dirichlet 0.6 | test accuracy | Mixed |
| Increase local training epochs from 30 to 40 (ablation) | DP-FedSAM · EMNIST, Dirichlet 0.6 | test accuracy | Hurt |
| Use more participating clients (ablation) | DP-FedSAM · EMNIST, Dirichlet 0.6 | test accuracy | Mixed |
| Use SAM rather than standard local optimization (ablation) | DP-FedAvg · EMNIST, Dirichlet 0.6 | validation accuracy (%): 82.20± 0.35 -> 84.32±0.19 | Helped |
Closest papers by meaning
- Towards the Robustness of Differentially Private Federated Learning (AAAI 2024 · similarity 0.91)
- Noise-Aware Algorithm for Heterogeneous Differentially Private Federated Learning (ICML 2024 · similarity 0.89)
- Mitigating the Privacy–Utility Trade-off in Decentralized Federated Learning via f-Differential Privacy (NeurIPS 2025 · similarity 0.88)
- DictPFL: Efficient and Private Federated Learning on Encrypted Gradients (NeurIPS 2025 · similarity 0.87)
- Enabling Differentially Private Federated Learning for Speech Recognition: Benchmarks, Adaptive Optimizers, and Gradient Clipping (NeurIPS 2025 · similarity 0.87)
- FedVSSAM: Mitigating Flatness Incompatibility in Sharpness-Aware Federated Learning (NeurIPS 2026 · 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.