Research map / Federated learning and privacy / Personalized and heterogeneous FL
PPFL: A Parameter Behavior-Driven Plug-in Personalization Engine for Federated Learning
Qianyue Cao, Zongwei Zhu, Zirui Lian, Rui Zhang, Boyu Li, Yi Xiong, Xuehai Zhou
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 parameter behavior awareness beyond update magnitude (ablation) | PPFL · Tiny-ImageNet, Dir alpha=0.1 | test accuracy (%): 49.59 -> 55.13 | Improved |
| add personalized fusion instead of simple averaging (ablation) | PPFL · Tiny-ImageNet, Dir alpha=0.1 | test accuracy (%): 51.24 -> 55.13 | Improved |
| increase local iterations from K=1 to K=10 (ablation) | PPFL · Tiny-ImageNet, Dir alpha=0.1 | test accuracy (%): 52.45 -> 55.82 | Improved |
| scale clients to 1000 with 20% participation (ablation) | PPFL · Tiny-ImageNet, Dir alpha=0.1 | test accuracy (%): 54.21 -> 59.74 | Improved |
| increase participation ratio from 0.2 to 1.0 at N=100 (ablation) | PPFL · Tiny-ImageNet, Dir alpha=0.1 | test accuracy (%): 54.21 -> 55.13 | Improved |
| use PPFL instead of FedDecomp | PPFL · Tiny-ImageNet, practical heterogeneity | local test accuracy (%): 52.89 -> 55.98 | Improved |
| use PPFL instead of FedDecomp | PPFL · Tiny-ImageNet, pathological heterogeneity | local test accuracy (%): 60.10 -> 65.43 | Improved |
| add PPFL plug-in to FedAvg | FedAvg · AG NEWS, alpha=3 | final local accuracy (%): 44.58 -> +35.13 | Improved |
| add PPFL plug-in to SCAFFOLD | SCAFFOLD · AG NEWS, alpha=3 | final local accuracy (%): 42.15 -> +14.07 | Improved |
| add PPFL plug-in to MOON | MOON · Tiny-ImageNet, alpha=3 | final local accuracy (%): 20.12 -> +14.89 | Improved |
| add PPFL fusion to FedAvg client drift | PPFL · Stitching loss barrier | max loss barrier: 14.76 -> 4.62 | Improved |
| add PPFL fusion to decoupled personalization | PPFL · Intrinsic drift barrier | loss barrier: 14.20 -> 4.37 | Improved |
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.
- FedAS — FedAS: Bridging Inconsistency in Personalized Federated Learning (CVPR 2024) · better on 2 of 2 shared results
- FedALA — FedALA: Adaptive Local Aggregation for Personalized Federated Learning (AAAI 2023) · better on 2 of 2 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
- PerAda: Parameter-Efficient Federated Learning Personalization with Generalization Guarantees (CVPR 2024 · similarity 0.90)
- Flexible Sharpness-Aware Personalized Federated Learning (AAAI 2025 · similarity 0.90)
- FedAS: Bridging Inconsistency in Personalized Federated Learning (CVPR 2024 · similarity 0.90)
- PeFLL: Personalized Federated Learning by Learning to Learn (ICLR 2024 · similarity 0.90)
- Understanding the Statistical Accuracy-Communication Trade-off in Personalized Federated Learning with Minimax Guarantees (ICML 2025 · similarity 0.89)
- Look Back for More: Harnessing Historical Sequential Updates for Personalized Federated Adapter Tuning (AAAI 2025 · similarity 0.88)
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.
