Research map / Federated learning and privacy / Personalized and heterogeneous FL
Fed-ADE: Adaptive Learning Rate for Federated Post-adaptation under Distribution Shift
Heewon Park, Mugon Joe, Miru Kim, Kyungjin Im, Minhae Kwon
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 cosine similarity instead of KL divergence as shift measure (ablation) | Fed-ADE · CIFAR-10 (label shift, Lin.) | accuracy (%): 63.2 -> 73.8 | Improved |
| use cosine similarity instead of Wasserstein distance (ablation) | Fed-ADE · CIFAR-10 (label shift, Lin.) | accuracy (%): 70.8 -> 73.8 | Improved |
| use cosine similarity instead of Bayesian CPD (ablation) | Fed-ADE · CIFAR-10 (label shift, Lin.) | accuracy (%): 71.5 -> 73.8 | Improved |
| add uncertainty dynamics estimator S_unc (ablation) | Fed-ADE · CIFAR-10 (label shift, Lin.) | accuracy (%): 71.3 -> 73.8 | Improved |
| add representation dynamics estimator S_rep (ablation) | Fed-ADE · CIFAR-10 (label shift, Lin.) | accuracy (%): 73.1 -> 73.8 | Improved |
| add uncertainty dynamics estimator S_unc (ablation) | Fed-ADE · CIFAR-10-C (covariate shift, Lin.) | accuracy (%): 64.2 -> 64.4 | Improved |
| add representation dynamics estimator S_rep (ablation) | Fed-ADE · CIFAR-10-C (covariate shift, Lin.) | accuracy (%): 64.0 -> 64.4 | Improved |
| replace fixed learning rate with adaptive learning rate from both estimators (ablation) | Fed-ADE · CIFAR-10 (label shift, Lin.) | accuracy (%): 70.8 -> 73.8 | Improved |
| replace fixed learning rate with adaptive learning rate from both estimators (ablation) | Fed-ADE · CIFAR-10-C (covariate shift, Lin.) | accuracy (%): 63.9 -> 64.4 | Improved |
| use federated adaptation with adaptive LR instead of localized adaptation (UNIDA) | Fed-ADE · CIFAR-10 (label shift, Lin.) | accuracy (%): 23.0 -> 73.8 | Improved |
| use Fed-ADE instead of Fed-POE under covariate shift | Fed-ADE · CIFAR-10-C (covariate shift, Lin.) | accuracy (%): 44.5 -> 64.4 | Improved |
| use Fed-ADE instead of Fed-POE under Gaussian pre-training distribution | Fed-ADE · CIFAR-10 (Gaussian pre-training, label shift, Lin.) | accuracy (%): 65.7 -> 69.2 | Improved |
Closest papers by meaning
- Federated Learning with Sample-level Client Drift Mitigation (AAAI 2025 · similarity 0.88)
- Efficient Federated Learning against Heterogeneous and Non-stationary Client Unavailability (NeurIPS 2024 · similarity 0.87)
- FedAdaVR: Adaptive Variance Reduction for Robust Federated Learning under Limited Client Participation (NeurIPS 2026 · similarity 0.87)
- FedImpro: Measuring and Improving Client Update in Federated Learning (ICLR 2024 · similarity 0.87)
- Adaptive Federated Learning with Auto-Tuned Clients (ICLR 2024 · similarity 0.87)
- Efficient Adaptive Federated Optimization (NeurIPS 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.
