Research map / Federated learning and privacy / Personalized and heterogeneous FL
Improving Generalization in Federated Learning by Seeking Flat Minima
Debora Caldarola, Barbara Caputo, Marco Ciccone
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 ASAM local training | FedAvg · CIFAR100 (alpha=0, 5 clients/round) | accuracy (%): 30.25 -> 36.04 | Improved |
| add SAM local training | FedAvg · CIFAR100 (alpha=0, 5 clients/round) | accuracy (%): 30.25 -> 31.04 | Improved |
| add ASAM local training | FedAvg · CIFAR10 (alpha=0, 5 clients/round) | accuracy (%): 65.00 -> 73.66 | Improved |
| add server-side SWA on top of ASAM | FedASAM · CIFAR100 (alpha=0, 5 clients/round) | accuracy (%): 36.04 -> 42.01 | Improved |
| add server-side SWA on top of ASAM | FedASAM · CIFAR10 (alpha=0, 5 clients/round) | accuracy (%): 73.66 -> 76.44 | Improved |
| add SWA to FedAvg | FedAvg · CIFAR100 (alpha=0, 5 clients/round) | accuracy (%): 30.25 -> 39.34 | Improved |
| add ASAM in homogeneous setting | FedAvg · CIFAR100 (alpha=1000, 5 clients/round) | accuracy (%): 49.92 -> 54.81 | Improved |
| use 2 local epochs instead of 1 | FedAvg · CIFAR100 (alpha=0, 5 clients/round) | accuracy (%): 30.25 -> 24.94 | Hurt |
| add ASAM local training (flatness) | FedAvg · CIFAR100 Hessian (alpha=0) | Hessian lambda_max: 93.46 -> 30.11 | Improved |
| add SWA on top of FedASAM (flatness) | FedASAM · CIFAR100 Hessian (alpha=0) | Hessian lambda_max: 30.11 -> 24.57 | Improved |
| add Mixup augmentation | FedAvg · CIFAR100 (alpha=0, 5 clients/round) | accuracy (%): 30.25 -> 29.91 | Hurt |
| add Cutout augmentation | FedAvg · CIFAR100 (alpha=0, 5 clients/round) | accuracy (%): 30.25 -> 24.24 | Hurt |
Closest papers by meaning
- Rethinking the Flat Minima Searching in Federated Learning (ICML 2024 · similarity 0.92)
- Locally Estimated Global Perturbations are Better than Local Perturbations for Federated Sharpness-aware Minimization (ICML 2024 · similarity 0.91)
- One Arrow, Two Hawks: Sharpness-aware Minimization for Federated Learning via Global Model Trajectory (ICML 2025 · similarity 0.91)
- Beyond Local Sharpness: Communication-Efficient Global Sharpness-aware Minimization for Federated Learning (CVPR 2025 · similarity 0.91)
- FedVSSAM: Mitigating Flatness Incompatibility in Sharpness-Aware Federated Learning (NeurIPS 2026 · similarity 0.91)
- Improving Generalization in Federated Learning with Highly Heterogeneous Data via Momentum-Based Stochastic Controlled Weight Averaging (ICML 2025 · similarity 0.90)
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.
