Research map / Federated learning and privacy / Personalized and heterogeneous FL
Few-for-Many Personalized Federated Learning
Ping Guo, Tiantian Zhang, Xi Lin, Xiang Li, Zhi-Ri Tang, Qingfu Zhang
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 |
|---|---|---|---|
| use FedFew instead of FedRep | FedFew · CIFAR-100 pathological M=20 | accuracy: 61.46 -> 64.98 | Improved |
| use FedFew instead of FedRep | FedFew · CIFAR-100 pathological M=10 | accuracy: 66.50 -> 65.47 | Hurt |
| use FedFew instead of strongest baseline | FedFew · CIFAR-100 practical M=20 | accuracy: 53.69% surpassing best baseline by 7.02% | Improved |
| use FedFew instead of FedRep | FedFew · TinyImageNet | accuracy: 27.24 -> 30.31 (+3.07) | Improved |
| use FedFew instead of FedRep | FedFew · AG News | accuracy: 94.68 -> 96.07 (+1.39) | Improved |
| use FedFew instead of APFL | FedFew · CIFAR-10 practical M=10 | accuracy: 88.38 -> 88.17 | Hurt |
| use FedFew instead of IFCA multi-model baseline | FedFew · FedISIC | average accuracy: 53.61 -> 69.57 | Improved |
| use FedFew instead of local-only training | FedFew · FedISIC minimum accuracy | min accuracy: +13.0% over local-only | Improved |
| increase number of server models K (ablation) | FedFew · CIFAR-10 (Dirichlet 0.5, M=20) | weighted accuracy: 89.4 to 91.3, K=1 highest at 91.3 | No clear effect |
| add multiple server models (K>1) versus single model (ablation) | FedFew · CIFAR-10 (Dirichlet 0.5, M=20) | weighted accuracy: 91.3 -> 89.4 to 91.3 | No clear effect |
| increase local epochs from 2 to 16 at fixed compute (ablation) | FedFew · CIFAR-10 communication | mean client accuracy: 88.3 -> 87.8 | No clear effect |
| increase K, convergence of STCH-Set objective (ablation) | FedFew · CIFAR-10 | convergence speed: convergence slows as K grows | 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.
- Ditto — Ditto: Fair and Robust Federated Learning Through Personalization (ICML 2021) · better on 1 of 1 shared result
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
- Understanding the Statistical Accuracy-Communication Trade-off in Personalized Federated Learning with Minimax Guarantees (ICML 2025 · similarity 0.91)
- pFedClub: Controllable Heterogeneous Model Aggregation for Personalized Federated Learning (NeurIPS 2024 · similarity 0.90)
- A Bayesian Approach for Personalized Federated Learning in Heterogeneous Settings (NeurIPS 2024 · similarity 0.90)
- Flexible Sharpness-Aware Personalized Federated Learning (AAAI 2025 · similarity 0.90)
- PeFLL: Personalized Federated Learning by Learning to Learn (ICLR 2024 · similarity 0.90)
- Fine-Tuning Personalization in Federated Learning to Mitigate Adversarial Clients (NeurIPS 2024 · similarity 0.90)
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.
