Research map / Federated learning and privacy / Personalized and heterogeneous FL
Dynamic Regularized Sharpness Aware Minimization in Federated Learning: Approaching Global Consistency and Smooth Landscape
Yan Sun, Li Shen, Shixiang Chen, Liang Ding, Dacheng Tao
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 global SAM perturbation with dynamic regularization (FedSMOO-style) versus dynamic regularization alone (ablation) | FedAvg base with dynamic regularization · CIFAR-10 Pathological c=3 | test accuracy (%): +2.50 (77.63 to 80.13) | Improved |
| Add SAM-correction (global perturbation correction) on top of dynamic regularization and SAM (ablation) | FedSMOO · CIFAR-10 Pathological c=3 | test accuracy (%): +1.45 (80.13 to 81.58) | Improved |
| Add dynamic regularization to FedAvg (ablation) | FedAvg · CIFAR-10 Pathological c=3 | test accuracy (%): +3.55 (74.08 to 77.63) | Improved |
| Full FedSMOO versus strongest baseline MoFedSAM | FedSMOO · CIFAR-10 Dirichlet u=0.1 | test accuracy (%): +2.11 over second-best | Improved |
| Full FedSMOO versus MoFedSAM on CIFAR-100 Dirichlet u=0.6 | FedSMOO · CIFAR-100 Dirichlet u=0.6 | test accuracy (%): 53.92 vs 54.38 for MoFedSAM | Hurt |
| Use SAM-based global perturbation (FedSAM) instead of FedAvg | FedSAM · CIFAR-10 Dirichlet u=0.1 | test accuracy (%): 76.86 vs 76.00 FedAvg | No clear effect |
| Add client-level momentum (MoFedSAM) to FedSAM | FedSAM · CIFAR-10 Dirichlet u=0.1 | test accuracy (%): 76.86 to 78.71 | Improved |
| Use FedSMOO instead of FedAvg in communication-rounds to 74% on CIFAR-10 Pathological-3 | FedSMOO · CIFAR-10 Pathological-3 | communication rounds to 74%: 194 vs 723 rounds (3.73x) | Improved |
Closest papers by meaning
- One Arrow, Two Hawks: Sharpness-aware Minimization for Federated Learning via Global Model Trajectory (ICML 2025 · similarity 0.94)
- Beyond Local Sharpness: Communication-Efficient Global Sharpness-aware Minimization for Federated Learning (CVPR 2025 · similarity 0.92)
- Locally Estimated Global Perturbations are Better than Local Perturbations for Federated Sharpness-aware Minimization (ICML 2024 · similarity 0.92)
- FedVSSAM: Mitigating Flatness Incompatibility in Sharpness-Aware Federated Learning (NeurIPS 2026 · similarity 0.91)
- Rethinking the Flat Minima Searching in Federated Learning (ICML 2024 · similarity 0.90)
- Flexible Sharpness-Aware Personalized Federated Learning (AAAI 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.
