Research map / Federated learning and privacy / Personalized and heterogeneous FL
Learning Locally, Revising Globally: Global Reviser for Federated Learning with Noisy Labels
Yuxin Tian, Mouxing Yang, Yuhao Zhou, Jian Wang, Qing Ye, Tongliang Liu, Gang Niu, Jiancheng Lv
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 federated sieving (global GMM on server over loss proxies) (ablation) | FedGR · CIFAR-10 Sym phi=1.0 I.I.D. | accuracy: 54.59 -> 83.91 | Improved |
| add federated sieving (ablation) | FedGR · CIFAR-10 Sym phi=1.0 Non-I.I.D. | accuracy: 45.48 -> 63.64 | Improved |
| add label refining (ablation) | FedGR · CIFAR-10 Sym phi=1.0 I.I.D. | accuracy: 75.23 -> 83.91 | Improved |
| add globally revised EMA distillation (ablation) | FedGR · CIFAR-10 Sym phi=1.0 Non-I.I.D. | accuracy: 51.07 -> 63.64 | Improved |
| add global representation regularization (ablation) | FedGR · CIFAR-10 Sym phi=1.0 I.I.D. | accuracy: 81.49 -> 83.91 | Improved |
| add weak augmentation on teacher/global inputs (ablation) | FedGR · CIFAR-10 Sym phi=1.0 Non-I.I.D. | accuracy: 25.32 -> 63.64 | Improved |
| add strong augmentation (ablation) | FedGR · CIFAR-10 Sym phi=1.0 I.I.D. | accuracy: 34.51 -> 83.91 | Improved |
| use globally revised EMA instead of online EMA distillation (ablation) | FedGR · CIFAR-10 Sym phi=1.0 I.I.D. | accuracy: 82.80 -> 83.91 | Improved |
| guarantee all clients sampled in warm-up (without replacement) (ablation) | FedGR · CIFAR-10 Sym phi=1.0 I.I.D. | accuracy: 82.56 -> 83.91 | Improved |
| increase lambda_R from 0.1 to 0.5 (ablation) | FedGR · CIFAR-10 Sym phi=1.0 I.I.D. | accuracy: 83.91 -> 52.74 | Hurt |
| increase gamma_g from 0.9 to 0.99 (ablation) | FedGR · CIFAR-10 Sym phi=1.0 I.I.D. | accuracy: 83.91 -> 79.87 | Hurt |
| increase gamma_l from 0.99 to 0.999 (ablation) | FedGR · CIFAR-10 | accuracy: 84.85 vs 83.91 (I.I.D. Sym); 50.86 vs 63.64 (Non-I.I.D. Sym) | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- FedCorr — FedCorr: Multi-Stage Federated Learning for Label Noise Correction (CVPR 2022) · better on 4 of 4 shared results
- FedFixer — FedFixer: Mitigating Heterogeneous Label Noise in Federated Learning (AAAI 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- FedDiv: Collaborative Noise Filtering for Federated Learning with Noisy Labels (AAAI 2024 · similarity 0.92)
- FedClean: A General Robust Label Noise Correction for Federated Learning (ICML 2025 · similarity 0.91)
- FedFixer: Mitigating Heterogeneous Label Noise in Federated Learning (AAAI 2024 · similarity 0.91)
- (FL)$^2$: Overcoming Few Labels in Federated Semi-Supervised Learning (NeurIPS 2024 · similarity 0.89)
- Federated Learning with Extremely Noisy Clients via Negative Distillation (AAAI 2024 · similarity 0.89)
- Subgraph Federated Learning for Local Generalization (ICLR 2025 · similarity 0.88)
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.
