Research map / Federated learning and privacy / Personalized and heterogeneous FL
Fed-GraB: Federated Long-tailed Learning with Self-Adjusting Gradient Balancer
Zikai Xiao, Zihan Chen, Songshang Liu, Hualiang Wang, YANG FENG, Jin Hao, Joey Tianyi Zhou, Jian Wu, Howard Hao Yang, Zuozhu Liu
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 Fed-GraB (DPA + SGB) to FedAvg | FedAvg · CIFAR-10-LT alpha=1, IF_G=100 | overall accuracy: 0.737 -> 0.766 | Improved |
| Add Fed-GraB to FedAvg | FedAvg · CIFAR-10-LT alpha=0.5, IF_G=100 | overall accuracy: 0.719 -> 0.761 | Improved |
| Add Fed-GraB to FedAvg | FedAvg · CIFAR-100-LT | overall accuracy: 0.365 -> 0.411 | Improved |
| Add Fed-GraB to FedAvg | FedAvg · ImageNet-LT | overall accuracy: 0.287 -> 0.311 | Improved |
| Add Fed-GraB to FedAvg | FedAvg · iNaturalist-160k | overall accuracy: 0.434 -> 0.451 | Improved |
| Apply SGB in centralized learning vs Softmax | Softmax · CIFAR-10-LT centralized, IF=100 | overall accuracy: 0.767 -> 0.818 | Improved |
| Use Fed-GraB instead of FedAvg for tail-class communication efficiency | FedAvg · CIFAR-10-LT IF_G=50 alpha=1 | rounds to 55% tail accuracy: 96 -> 59 | Improved |
| Use Fed-GraB instead of FedAvg for tail-class communication efficiency | FedAvg · CIFAR-10-LT IF_G=100 alpha=0.5 | rounds to 55% tail accuracy: 192 -> 122 | Improved |
| Use Fed-GraB instead of Eqlv2-FL, which cannot align Delta(t) | Eqlv2-FL · CIFAR-10-LT IF_G=50 alpha=1 | rounds to 55% tail accuracy: 301 -> 59 | Improved |
| Use global prior vector from DPA instead of local distribution (Local-SGB) (ablation) | SGB · CIFAR-10-LT | test accuracy | Improved |
| Mount SGB on all classes with DPA instead of only tail classes 7 to 9 (ablation) | SGB · CIFAR-10-LT | test accuracy | Improved |
| Use proportional gain K_P=10 versus K_P=1 (ablation) | SGB controller · CIFAR-10-LT | Delta(t) control | Improved |
Later papers that compared with this method
1 later accepted paper reports Fed-Grab as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- You Are Your Own Best Teacher: Achieving Centralized-level Performance in Federated Learning under Heterogeneous and Long-tailed Data (ICCV 2025 · better on 4 of 4 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
- FedLoGe: Joint Local and Generic Federated Learning under Long-tailed Data (ICLR 2024 · similarity 0.89)
- Federated Learning with Sample-level Client Drift Mitigation (AAAI 2025 · similarity 0.86)
- FedRTS: Federated Robust Pruning via Combinatorial Thompson Sampling (NeurIPS 2025 · similarity 0.86)
- FedImpro: Measuring and Improving Client Update in Federated Learning (ICLR 2024 · similarity 0.86)
- FedAWA: Adaptive Optimization of Aggregation Weights in Federated Learning Using Client Vectors (CVPR 2025 · similarity 0.86)
- FedGPS: Statistical Rectification Against Data Heterogeneity in Federated Learning (NeurIPS 2025 · 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.
