Research map / Federated learning and privacy / Personalized and heterogeneous FL
HFedATM: Hierarchical Federated Domain Generalization via Optimal Transport and Regularized Mean Aggregation
Thinh Nguyen, Trung Phan, Binh Nguyen, Khoa D Doan, Kok-Seng Wong
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 |
|---|---|---|---|
| replace station-server averaging with HFedATM | FedIIR · PACS avg accuracy (ResNet-18) | accuracy: 89.7 -> 93.0 | Improved |
| replace station-server averaging with HFedATM | FedSR · PACS target Photo (lambda=1.0) | accuracy: 84.1 -> 87.7 | Improved |
| replace station-server averaging with HFedATM | FedAvg · Amazon Reviews target Books (lambda=1.0) | accuracy: 70.3 -> 78.1 | Improved |
| replace station-server averaging with HFedATM under strong heterogeneity | FedAvg · TerraInc target L38 (lambda=1.0) | accuracy: 45.7 -> 45.5 | No clear effect |
| replace averaging with HFedATM when client models are weakly DG and heterogeneous | FedAvg and FedProx · Office-Home and TerraInc (lambda <= 0.1) | accuracy | Mixed |
| add FOT alignment (ablation) | FedAvg + HFedATM · PACS (lambda=1.0) | accuracy: 76.9 -> 78.3 | Improved |
| add RegMean aggregation (ablation) | FedAvg + HFedATM · PACS (lambda=1.0) | accuracy: 77.7 -> 78.3 | Improved |
| add FOT alignment (ablation) | FedSR + HFedATM · Amazon Reviews (lambda=1.0) | accuracy: 73.0 -> 81.0 | Improved |
| add RegMean aggregation (ablation) | FedProx + HFedATM · TerraInc (lambda=1.0) | accuracy: 41.7 -> 44.2 | Improved |
| add HFedATM aggregation to FedAvg | FedAvg · PACS training time per round | training time (s): 20.4 -> 22.1 | Hurt |
| apply HFedATM across backbones (VGG-11) | FedSR · PACS avg accuracy (VGG-11) | accuracy: 88.3 -> 91.7 | Improved |
| add Gaussian DP noise to shared Gram matrices (epsilon >= 1) (ablation) | HFedATM · PACS, Office-Home, TerraInc, Amazon Reviews | accuracy: drop of less than 2% | No clear effect |
Closest papers by meaning
- Hierarchical Federated Learning with Multi-Timescale Gradient Correction (NeurIPS 2024 · similarity 0.90)
- Benchmarking Algorithms for Federated Domain Generalization (ICLR 2024 · similarity 0.89)
- Federated Domain Generalization with Data-free On-server Matching Gradient (ICLR 2025 · similarity 0.89)
- FedQS: Optimizing Gradient and Model Aggregation for Semi-Asynchronous Federated Learning (NeurIPS 2025 · similarity 0.88)
- FedGPS: Statistical Rectification Against Data Heterogeneity in Federated Learning (NeurIPS 2025 · similarity 0.88)
- TTA-FedDG: Leveraging Test-Time Adaptation to Address Federated Domain Generalization (AAAI 2025 · 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.
