Research map / Federated learning and privacy / Personalized and heterogeneous FL
You Are Your Own Best Teacher: Achieving Centralized-level Performance in Federated Learning under Heterogeneous and Long-tailed Data
Shanshan Yan, Zexi Li, Chao Wu, Meng Pang, Yang Lu, Yan Yan, Hanzi Wang
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 ASD on top of RandAug (ablation) | FedYoYo · CIFAR-100-LT (IF=100, alpha=0.5) | top-1 accuracy: accuracy improves by 8.65% | Improved |
| add DLA on top of RandAug and ASD (ablation) | FedYoYo · CIFAR-100-LT (IF=100, alpha=0.5) | top-1 accuracy: 41.99 -> 46.13 | Improved |
| add ASD on top of RandAug and DLA (ablation) | FedYoYo · CIFAR-100-LT (IF=100, alpha=0.5) | top-1 accuracy: 41.69 -> 46.13 | Improved |
| add RandAug alone to vanilla model (ablation) | FedYoYo · CIFAR-100-LT (IF=100, alpha=0.5) | top-1 accuracy: 33.34 -> 35.71 | Improved |
| add DLA to RandAug+ASD for few-shot classes (ablation) | FedYoYo · CIFAR-100-LT (IF=100, alpha=0.5) | few-shot accuracy: 13.93 -> 28.63 | Improved |
| add DLA to RandAug+ASD for many-shot classes (ablation) | FedYoYo · CIFAR-100-LT (IF=100, alpha=0.5) | many-shot accuracy: 65.31 -> 60.26 | Hurt |
| weak teacher / strong student instead of strong/strong views (ablation) | FedYoYo · CIFAR-100-LT (IF=100, alpha=0.5) | top-1 accuracy: 42.92 -> 46.13 | Improved |
| weak teacher / strong student instead of reversed roles (ablation) | FedYoYo · CIFAR-100-LT (IF=100, alpha=0.5) | top-1 accuracy: 41.06 -> 46.13 | Improved |
| weak teacher / strong student instead of weak/weak (ablation) | FedYoYo · CIFAR-100-LT (IF=100, alpha=0.5) | top-1 accuracy: 33.74 -> 46.13 | Improved |
| use AutoAugment or TrivialAugment instead of RandAugment (ablation) | FedYoYo · CIFAR-100-LT (IF=100) | top-1 accuracy: 46.13 vs 46.01 vs 46.11 | No clear effect |
| fuse local and global distribution (gamma) rather than gamma=1.0 (ablation) | FedYoYo · CIFAR-100-LT | top-1 accuracy | Improved |
| increase lambda from 3.0 to 5.5 (ablation) | FedYoYo · CIFAR-100-LT | top-1 accuracy | No clear effect |
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.
- Fed-Grab — Fed-GraB: Federated Long-tailed Learning with Self-Adjusting Gradient Balancer (NeurIPS 2023) · better on 4 of 4 shared results
- FedLoGe — FedLoGe: Joint Local and Generic Federated Learning under Long-tailed Data (ICLR 2024) · better on 3 of 3 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 0.91)
- FedUV: Uniformity and Variance for Heterogeneous Federated Learning (CVPR 2024 · similarity 0.87)
- FedQS: Optimizing Gradient and Model Aggregation for Semi-Asynchronous Federated Learning (NeurIPS 2025 · similarity 0.87)
- Widening the Network Mitigates the Impact of Data Heterogeneity on FedAvg (ICML 2025 · similarity 0.87)
- FedSMU: Communication-Efficient and Generalization-Enhanced Federated Learning through Symbolic Model Updates (ICML 2025 · similarity 0.87)
- Rethinking the Representation in Federated Unsupervised Learning with Non-IID Data (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.
