Research map / Federated learning and privacy / Communication-efficient FL
Federated Hyperparameter Tuning: Challenges, Baselines, and Connections to Weight-Sharing
Mikhail Khodak, Renbo Tu, Tian Li, Liam Li, Nina Balcan, Virginia Smith, Ameet Talwalkar
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 |
|---|---|---|---|
| Tune client hyperparameters with FedEx rather than tuning server and client hyperparameters with SHA alone | Global model · Shakespeare i.i.d. | test error: 47.38 \pm 3.40 -> 44.52 \pm 1.68 | Improved |
| Tune client hyperparameters with FedEx rather than tuning server and client hyperparameters with SHA alone | Global model · Shakespeare non-i.i.d. | test error: 46.79 \pm 3.51 -> 45.24 \pm 3.31 | Improved |
| Tune client hyperparameters with FedEx rather than tuning server and client hyperparameters with SHA alone | Global model · FEMNIST i.i.d. | test error: 18.64 \pm 1.68 -> 19.22 \pm 2.05 | Hurt |
| Tune client hyperparameters with FedEx rather than tuning server and client hyperparameters with SHA alone | Global model · FEMNIST non-i.i.d. | test error: 20.30 \pm 1.66 -> 19.43 \pm 1.45 | Improved |
| Tune client hyperparameters with FedEx rather than tuning server and client hyperparameters with SHA alone | Global model · CIFAR-10 i.i.d. | test error: 21.62 \pm 2.51 -> 20.82 \pm 1.37 | Improved |
| Tune client hyperparameters with FedEx rather than tuning server and client hyperparameters with SHA alone | Personalized model · Shakespeare i.i.d. | test error: 46.77 \pm 3.61 -> 46.08 \pm 2.57 | Improved |
| Tune client hyperparameters with FedEx rather than tuning server and client hyperparameters with SHA alone | Personalized model · Shakespeare non-i.i.d. | test error: 48.04 \pm 3.72 -> 45.89 \pm 3.76 | Improved |
| Tune client hyperparameters with FedEx rather than tuning server and client hyperparameters with SHA alone | Personalized model · FEMNIST i.i.d. | test error: 14.79 \pm 1.55 -> 14.97 \pm 1.31 | Hurt |
| Tune client hyperparameters with FedEx rather than tuning server and client hyperparameters with SHA alone | Personalized model · FEMNIST non-i.i.d. | test error: 14.78 \pm 1.31 -> 14.76 \pm 1.70 | Improved |
| Tune client hyperparameters with FedEx rather than tuning server and client hyperparameters with SHA alone | Personalized model · CIFAR-10 i.i.d. | test error: 24.81 \pm 6.13 -> 21.77 \pm 2.83 | Improved |
| Use nonzero local perturbation instead of setting ε=0 (SHA only) (ablation) | FedEx · CIFAR-10 | test error | Improved |
| Use fairly low local perturbation ε rather than independently sampling configurations (ε=1.0) (ablation) | FedEx · CIFAR-10 | test error | No clear effect |
Closest papers by meaning
- FedPop: Federated Population-based Hyperparameter Tuning (AAAI 2025 · similarity 0.91)
- Adaptive Federated Learning with Auto-Tuned Clients (ICLR 2024 · similarity 0.88)
- FedHyper: A Universal and Robust Learning Rate Scheduler for Federated Learning with Hypergradient Descent (ICLR 2024 · similarity 0.87)
- Federated Optimization with Doubly Regularized Drift Correction (ICML 2024 · similarity 0.86)
- Locally Estimated Global Perturbations are Better than Local Perturbations for Federated Sharpness-aware Minimization (ICML 2024 · similarity 0.86)
- Probabilistic Federated Prompt-Tuning with Non-IID and Imbalanced Data (NeurIPS 2024 · 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.
