Research map / Federated learning and privacy / Personalized and heterogeneous FL
Dynamic Personalized Federated Learning with Adaptive Differential Privacy
Xiyuan Yang, Wenke Huang, Mang Ye
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 |
|---|---|---|---|
| add Dynamic Fisher Personalization (ablation) | DP-FedAvg base · FEMNIST eps=2 | average client accuracy (%): 71.92 -> 72.18 | Improved |
| add Dynamic Fisher Personalization (ablation) | DP-FedAvg base · FEMNIST eps=4 | average client accuracy (%): 72.80 -> 74.56 | Improved |
| add Adaptive Constraint on top of DFP (ablation) | FedDPA · FEMNIST eps=2 | average client accuracy (%): 72.18 -> 74.46 | Improved |
| add Adaptive Constraint on top of DFP (ablation) | FedDPA · FEMNIST eps=16 | average client accuracy (%): 75.33 -> 76.99 | Improved |
| personalize by Fisher mask threshold tau=0.2 instead of tau=0 (all parameters personalized) (ablation) | FedDPA · hyperparameter study | accuracy (%): 39.72 -> 49.28 | Improved |
| vary tau between 0.1 and 1 (ablation) | FedDPA · hyperparameter study | accuracy (%): 47.97 - 49.28 | No clear effect |
| add L2 constraint on personalized updates lambda1=0.05 (ablation) | FedDPA · hyperparameter study | accuracy (%): 60.82 -> 62.64 | Improved |
| increase lambda1 from 0.05 to 0.1 or more (ablation) | FedDPA · hyperparameter study | accuracy (%): 62.64 -> 59.66 | Hurt |
| add bounded regularization on shared updates lambda2=0.1 (ablation) | FedDPA · hyperparameter study | accuracy (%): 58.34 -> 61.01 | Improved |
| increase lambda2 to 0.3 or 0.5 (ablation) | FedDPA · hyperparameter study | accuracy (%): 61.01 -> 45.66 | Hurt |
| FedDPA vs best baseline | FedDPA · SVHN eps=6 | average client accuracy (%): +6.76 | Improved |
Closest papers by meaning
- Personalization Improves Privacy-Accuracy Tradeoffs in Federated Learning (ICML 2022 · similarity 0.90)
- A Statistical Framework for Personalized Federated Learning and Estimation: Theory, Algorithms, and Privacy (ICLR 2023 · similarity 0.89)
- A Bayesian Approach for Personalized Federated Learning in Heterogeneous Settings (NeurIPS 2024 · similarity 0.87)
- Private Model Personalization Revisited (ICML 2025 · similarity 0.87)
- Noise-Aware Algorithm for Heterogeneous Differentially Private Federated Learning (ICML 2024 · similarity 0.86)
- Mitigating the Privacy–Utility Trade-off in Decentralized Federated Learning via f-Differential Privacy (NeurIPS 2025 · similarity 0.86)
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.
