Research map / Federated learning and privacy / Personalized and heterogeneous FL
Towards Personalized Federated Learning via Heterogeneous Model Reassembly
Jiaqi Wang, Xingyi Yang, Suhan Cui, Liwei Che, Lingjuan Lyu, Dongkuan Xu, Fenglong Ma
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 pFedHR instead of best baseline (FedGH) with labeled public data | heterogeneous FL · SVHN Non-IID | accuracy: 81.06% -> 83.40% | Improved |
| use pFedHR instead of FCCL with unlabeled public data | heterogeneous FL · CIFAR-10 IID | accuracy: 68.77% -> 69.38% | Improved |
| use pFedHR instead of FedMD with 100 clients | heterogeneous FL · SVHN Non-IID, N=100 | accuracy: 74.34% -> 77.63% | Improved |
| use pFedHR with homogeneous clients vs best personalized baseline, simple model M1 | homogeneous FL · CIFAR-10 IID | accuracy: 46.12% -> 54.38% | Improved |
| use pFedHR with homogeneous clients on MNIST | homogeneous FL · MNIST IID, M4 | accuracy: 95.77% -> 94.88% | Hurt |
| use labeled instead of unlabeled public data (ablation) | pFedHR · CIFAR-10 IID | accuracy: 69.38% -> 73.88% | Improved |
| increase number of stitching linear layers from 1 to 3 (ablation) | pFedHR · SVHN, 12 clients | accuracy | Hurt |
| increase server finetuning epochs of stitched model (ablation) | pFedHR · SVHN, 12 clients | accuracy | Hurt |
| increase percentage of public data from 2% to 5% to 10% (ablation) | pFedHR · SVHN | accuracy | Improved |
| increase number of clusters K (ablation) | pFedHR · SVHN, 12 clients | accuracy | Improved |
| use different-distribution public data instead of SVHN public data (ablation) | pFedHR and baselines · SVHN | accuracy | Hurt |
Closest papers by meaning
- pFedClub: Controllable Heterogeneous Model Aggregation for Personalized Federated Learning (NeurIPS 2024 · similarity 0.93)
- Look Back for More: Harnessing Historical Sequential Updates for Personalized Federated Adapter Tuning (AAAI 2025 · similarity 0.89)
- pFedES: Generalized Proxy Feature Extractor Sharing for Model Heterogeneous Personalized Federated Learning (AAAI 2025 · similarity 0.89)
- FedGPS: Statistical Rectification Against Data Heterogeneity in Federated Learning (NeurIPS 2025 · similarity 0.89)
- FedAS: Bridging Inconsistency in Personalized Federated Learning (CVPR 2024 · similarity 0.88)
- Decentralized Directed Collaboration for Personalized Federated Learning (CVPR 2024 · 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.
