Research map / Federated learning and privacy / Personalized and heterogeneous FL
Factorized-FL: Personalized Federated Learning with Parameter Factorization & Similarity Matching
Wonyong Jeong, Sung Ju Hwang
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 |
|---|---|---|---|
| aggregate all parameters (beta) instead of only basis vectors (alpha) | Factorized-FL · CIFAR-10 Standard Non-IID | accuracy %: 50.34 -> 65.30 | Improved |
| beta vs alpha on domain-heterogeneous average | Factorized-FL · domain & label heterogeneous | accuracy %: 64.49 -> 63.93 | Hurt |
| beta vs alpha communication cost, domain-heterogeneous | Factorized-FL · domain & label heterogeneous | cost Gb: 0.64 -> 36.5 | Hurt |
| permute labels across clients | FedAvg · CIFAR-10 Standard IID | accuracy %: 70.28 -> 65.31 | Hurt |
| permute labels across clients | Factorized-FL alpha · CIFAR-10 IID | accuracy %: 66.97 -> 67.91 | No clear effect |
| naive aggregation in domain-heterogeneous setting | FedAvg vs Stand-Alone · domain & label heterogeneous | accuracy %: 61.35 -> 56.42 | Hurt |
| kernel factorization (Factorized-FedAvg vs FedAvg) (ablation) | FedAvg · domain-heterogeneous | accuracy | Improved |
| remove sparse bias matrix M (ablation) | Factorized-FedAvg · domain-heterogeneous | accuracy | Hurt |
| factorize u in R^{F.F}, v in R^{I.O} instead of u in R^{I.F}, v in R^{O.F} (ablation) | Factorized-FL · domain-heterogeneous | accuracy | Improved |
| random or worst matching instead of similarity matching (ablation) | Factorized-FL · multi-domain | accuracy | Hurt |
| increase gamma (rank) | Factorized-FL · domain-heterogeneous | accuracy | Hurt |
| lower similarity threshold tau | Factorized-FL · domain-heterogeneous | accuracy | Hurt |
Closest papers by meaning
- PFedCS: A Personalized Federated Learning Method for Enhancing Collaboration among Similar Classifiers (AAAI 2025 · similarity 0.89)
- FedAS: Bridging Inconsistency in Personalized Federated Learning (CVPR 2024 · similarity 0.88)
- pFedClub: Controllable Heterogeneous Model Aggregation for Personalized Federated Learning (NeurIPS 2024 · similarity 0.88)
- PeFLL: Personalized Federated Learning by Learning to Learn (ICLR 2024 · similarity 0.87)
- Adaptive Latent-Space Constraints in Personalized Federated Learning (NeurIPS 2025 · similarity 0.87)
- A Swiss Army Knife for Heterogeneous Federated Learning: Flexible Coupling via Trace Norm (NeurIPS 2024 · similarity 0.87)
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.
