Research map / Federated learning and privacy / Personalized and heterogeneous FL
FedCART: Tackling Long-Tailed Distributions in Federated Adversarial Training via Classifier Refinement
Yuchen Qin, Yizhi Zhou, Junxiao Wang, Xin Xie, Heng Qi
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 alignment loss LAlign between natural and adversarial features (ablation) | FedCART · CIFAR10-LT | robust accuracy PGD-20: 29.33 -> 32.57 | Improved |
| add adversarial loss LAT (ablation) | FedCART · CIFAR10-LT | robust accuracy PGD-20: 2.91 -> 32.57 | Improved |
| add server-side classifier re-training with auxiliary classifier (ablation) | FedCART · CIFAR10-LT | natural accuracy: 49.98 -> 67.24 | Improved |
| replace CalFAT with FedCART under long-tailed data | CalFAT · CIFAR10-LT | robust accuracy AA: 19.93 -> 27.47 | Improved |
| FedCART vs CalFAT under small beta (strong heterogeneity) | FedCART · CIFAR10-LT (beta=0.25) | natural accuracy: 55.65 -> 51.52 | Hurt |
| FedCART vs CalFAT under small beta (strong heterogeneity) | FedCART · CIFAR10-LT (beta=0.25) | robust accuracy AA: 18.00 -> 24.72 | Improved |
| FedCART vs CalFAT on FMNIST-LT natural accuracy | FedCART · FMNIST-LT | natural accuracy: 86.75 -> 85.13 | Hurt |
| increase imbalance ratio rho | CalFAT and FedCART · CIFAR10-LT | accuracy | Hurt |
| combine TRADES with FedCART | FedTRADES · CIFAR10-LT | natural accuracy: 41.59 -> 65.48 | Improved |
| combine TAET with FedCART | FedTAET · CIFAR10-LT | robust accuracy PGD-20: 22.42 -> 26.57 | Improved |
| FedCART on majority classes vs FedFAT | FedCART · CIFAR100-LT majority classes | natural accuracy: 60.45 -> 52.16 | Hurt |
| decrease client active rate or increase client count | CalFAT and FedCART · CIFAR10-LT | accuracy | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CalFAT — CalFAT: Calibrated Federated Adversarial Training with Label Skewness (NeurIPS 2022) · better on 5 of 7 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
- CalFAT: Calibrated Federated Adversarial Training with Label Skewness (NeurIPS 2022 · similarity 0.89)
- FedLoGe: Joint Local and Generic Federated Learning under Long-tailed Data (ICLR 2024 · similarity 0.88)
- Byzantine-Robust Federated Learning: Impact of Client Subsampling and Local Updates (ICML 2024 · similarity 0.88)
- TAET: Two-Stage Adversarial Equalization Training on Long-Tailed Distributions (CVPR 2025 · similarity 0.87)
- Long-tailed Adversarial Training with Self-Distillation (ICLR 2025 · similarity 0.87)
- Revisiting Adversarial Training Under Long-Tailed Distributions (CVPR 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.
