Research map / Federated learning and privacy / Personalized and heterogeneous FL
FedRNC: Addressing Spatio-Temporal Label Misalignment in Federated Noisy Class-Incremental Learning
Xingwei Huang, Zhaobin Sun, Junjie Shi, Xin Yang, Zengqiang Yan
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 prototype-guided relabeling (stage 2) after loss filtering (ablation) | FedRNC · Tiny-ImageNet IID noise (rho=40%) | BACC (%): 34.8 -> 40.4 | Improved |
| add initial local loss filtering (stage 1) (ablation) | FedRNC · Tiny-ImageNet IID noise (rho=40%) | BACC (%): 32.6 -> 40.4 | Improved |
| add task-wise similarity grouping in prototype matching (ablation) | FedRNC · Tiny-ImageNet IID noise (rho=40%) | BACC (%): 36.1 -> 40.4 | Improved |
| add reuse and relabeling of noisy samples (ablation) | FedRNC · Tiny-ImageNet IID noise (rho=40%) | BACC (%): 36.8 -> 40.4 | Improved |
| relabel noisy samples instead of simply excluding them (ablation) | FedRNC · Tiny-ImageNet IID noise (rho=40%) | BACC (%): 35.7 -> 40.4 | Improved |
| add both FedRNC stages (vs removing prototype matching) to reduce forgetting (ablation) | FedRNC · Tiny-ImageNet IID noise (rho=40%) | forgetting: 9.2 -> 5.8 | Improved |
| plug FedRNC denoising into FL+iCaRL | FL+iCaRL · EMNIST (rho=60%) | accuracy (%): 40.3 -> 56.3 | Improved |
| plug FedRNC denoising into GLFC | GLFC · EMNIST (rho=60%) | accuracy (%): 53.7 -> 65.8 | Improved |
| plug FedRNC denoising into Re-Fed | Re-Fed · EMNIST (rho=20%) | accuracy (%): 65.0 -> 70.9 | Improved |
| use FedRNC rather than best baseline at lowest noise on CIFAR-100 | FedRNC · CIFAR-100 FNCIL IID-noise (rho=20%) | accuracy (%): 46.7 -> 45.7 | Hurt |
| increase noise rate from 20% to 60% for FedRNC on EMNIST | FedRNC · EMNIST FNCIL IID-noise | accuracy (%): 71.3 -> 71.2 | No clear effect |
Closest papers by meaning
- FedClean: A General Robust Label Noise Correction for Federated Learning (ICML 2025 · similarity 0.88)
- FedDiv: Collaborative Noise Filtering for Federated Learning with Noisy Labels (AAAI 2024 · similarity 0.87)
- Federated Partial Label Learning with Local-Adaptive Augmentation and Regularization (AAAI 2024 · similarity 0.86)
- Exploring Vacant Classes in Label-Skewed Federated Learning (AAAI 2025 · similarity 0.86)
- FedFixer: Mitigating Heterogeneous Label Noise in Federated Learning (AAAI 2024 · similarity 0.85)
- Text-Enhanced Data-free Approach for Federated Class-Incremental Learning (CVPR 2024 · similarity 0.85)
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.
