Research map / Federated learning and privacy / Personalized and heterogeneous FL
PartialFed: Cross-Domain Personalized Federated Learning via Partial Initialization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| load only part of global params (keep others local) instead of full global init (ablation) | FedAvg with ResNet-18 · Office-Home | mean accuracy: 75.44 -> 78.50 (s1) up to 80.32 | Helped |
| keep BN layers local (ablation) | FedAvg with ResNet-18 · Office-Home | mean accuracy: 75.44 -> 78.78 | Helped |
| keep BN and fc local (ablation) | FedAvg with ResNet-18 · Office-Home | mean accuracy: 75.44 -> 80.32 | Helped |
| load global fc layer only (share classifier) rather than keep fc local (ablation) | PartialFed-Fix · Office-Home | mean accuracy: 78.86 -> 78.73 (s1-s4 vs fc); fc-including s1-fc = 75.44 | Hurt |
| share more top-level stages globally (s2-fc vs fc) (ablation) | PartialFed-Fix · Office-Home | mean accuracy: 78.73 -> 77.63 | Hurt |
| skip loading (alternating global/local blocks) vs adjacent block loading (ablation) | PartialFed-Fix · Office-Home | mean accuracy: 78.66 -> 79.31 (AaBb vs ABCD) | Helped |
| learn loading strategy adaptively (layer-level) instead of fixed (ablation) | PartialFed · Office-Home | mean accuracy: 80.32 -> 80.87 | Helped |
| layer-level instead of stage-level adaptive granularity (ablation) | PartialFed-Adaptive · Office-Home | mean accuracy: 80.46 -> 80.87 | Helped |
| layer-level instead of block-level adaptive granularity (ablation) | PartialFed-Adaptive · UODB | mean AP50: 64.02 -> 63.53 | Hurt |
| PartialFed-Adaptive block vs FedAvg | FedAvg · UODB | mean AP50: 61.34 -> 64.02 | Helped |
| FedBN vs FedAvg on DeepLesion | FedAvg · UODB DeepLesion | AP50: 58.26 -> 57.11 | Hurt |
| add FedProx regularization | FedAvg · UODB | mean AP50: 61.34 -> 60.71 | Hurt |
Closest papers by meaning
- Dual-Personalizing Adapter for Federated Foundation Models (NeurIPS 2024 · similarity 0.87)
- pFedES: Generalized Proxy Feature Extractor Sharing for Model Heterogeneous Personalized Federated Learning (AAAI 2025 · similarity 0.86)
- Adaptive Latent-Space Constraints in Personalized Federated Learning (NeurIPS 2025 · similarity 0.86)
- pFedClub: Controllable Heterogeneous Model Aggregation for Personalized Federated Learning (NeurIPS 2024 · similarity 0.86)
- pFedMxF: Personalized Federated Class-Incremental Learning with Mixture of Frequency Aggregation (CVPR 2025 · similarity 0.86)
- Decentralized Directed Collaboration for Personalized Federated Learning (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.