Research map / Federated learning and privacy / Personalized and heterogeneous FL
Distribution Alignment for One-Shot Federated Learning via Optimal Transport
Daniele Berardini, Vito Paolo Pastore, Vittorio Murino
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 SLOT-Align feature alignment | O-FedAvg · Office-Home (alpha=0.1) | macro accuracy: 64.19 -> 72.05 | Improved |
| add SLOT-Align feature alignment | FedPFT · Office-Home (alpha=0.1) | macro accuracy: 74.19 -> 80.74 | Improved |
| add SLOT-Align feature alignment | FedCGS · Office-Home (alpha=0.1) | macro accuracy: 84.37 -> 85.05 | Improved |
| add SLOT-Align feature alignment | FedCGS · DomainNet (alpha=0.1) | macro accuracy: 51.40 -> 56.63 | Improved |
| add SLOT-Align feature alignment | FedPFT · Digits (alpha=0.01) | macro accuracy: 43.69 -> 50.98 | Improved |
| add SLOT-Align feature alignment | FedPFT · Office-Home with ResNet-18 backbone | macro accuracy: 48.22 -> 50.62 | Improved |
| add SLOT-Align when labels are balanced (ablation) | O-FedAvg · Office-Home (domain shift only) | macro accuracy: 90.94 -> 91.00 | No clear effect |
| add SLOT-Align when labels are balanced (ablation) | FedPFT · Office-Home (domain shift only) | macro accuracy: 90.22 -> 90.17 | No clear effect |
| raise alignment strength tau from 0.4 to 1.0 (full OT map) (ablation) | O-FedAvg + SLOT-Align · Office-Home (alpha=0.1) | macro accuracy: 72.05 -> 69.42 | Hurt |
| raise alignment strength tau from 0.4 to 1.0 (full OT map) (ablation) | FedCGS + SLOT-Align · Office-Home (alpha=0.1) | macro accuracy: 85.05 -> 80.05 | Hurt |
| use moderate alignment strength tau=0.4 versus none (ablation) | FedPFT + SLOT-Align · Office-Home (alpha=0.1) | macro accuracy: 74.19 -> 80.74 | Improved |
| use very small alignment strength tau=0.2 versus none (ablation) | FedCGS + SLOT-Align · Office-Home (alpha=0.1) | macro accuracy: 84.37 -> 84.40 | No clear effect |
Closest papers by meaning
- Geometric Knowledge-Guided Localized Global Distribution Alignment for Federated Learning (CVPR 2025 · similarity 0.87)
- Revisiting Ensembling in One-Shot Federated Learning (NeurIPS 2024 · similarity 0.85)
- Capture Global Feature Statistics for One-Shot Federated Learning (AAAI 2025 · similarity 0.85)
- FedGPS: Statistical Rectification Against Data Heterogeneity in Federated Learning (NeurIPS 2025 · similarity 0.85)
- FedBEns: One-Shot Federated Learning based on Bayesian Ensemble (ICML 2025 · similarity 0.85)
- FedLPA: One-shot Federated Learning with Layer-Wise Posterior Aggregation (NeurIPS 2024 · similarity 0.85)
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.
