Research map / Federated learning and privacy / Communication-efficient FL
Federated Gaussian Process: Convergence, Automatic Personalization and Multi-Fidelity Modeling
Xubo Yue, Raed Al Kontar
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 FGPR (federated global prior) instead of separate HF-only GP | FGPR · CURRIN | RMSE vs true HF model: 0.301 -> 0.148 | Improved |
| use FGPR instead of centralized Deep multi-fidelity model | FGPR · PARK | RMSE vs true HF model: 0.013 -> 0.012 | Improved |
| use FGPR instead of centralized Deep multi-fidelity model | FGPR · BRANIN | RMSE vs true HF model: 0.213 -> 0.260 | Hurt |
| use FGPR instead of AR multi-fidelity model | FGPR · Borehole | RMSE vs true HF model: 0.615 -> 0.604 | Improved |
| increase HF client weight p1 from 0.17 toward 1 (decrease LF weight) (ablation) | FGPR · CURRIN | RMSE: 0.17 to 1 | Mixed |
| use FGPR instead of federated polynomial regression | FGPR · NASA C-MAPSS Sensor 2 | averaged RMSE x10: 6.79 -> 5.45 | Improved |
| use FGPR instead of federated neural network | FGPR · NASA C-MAPSS Sensor 7 | averaged RMSE x10: 6.71 -> 5.76 | Improved |
| use FGPR instead of Ditto-personalized neural network | FGPR · SARCOS Output 5 | averaged RMSE x10: 2.89 -> 2.20 | Improved |
| use FGPR instead of Ditto-personalized neural network | FGPR · SARCOS Output 1 | std of RMSE x10: 1.70 -> 1.84 | Hurt |
| start FGPR from a bad initialization (flat-line fit, theta=(1,10,1)) (ablation) | FGPR · sin(x) toy, two clients | averaged RMSE | No clear effect |
| use few data points per device (K=100, N=800) (ablation) | FGPR · Homogeneous GP simulation | convergence: N=800, M_k=8 | Mixed |
Closest papers by meaning
- FedLMT: Tackling System Heterogeneity of Federated Learning via Low-Rank Model Training with Theoretical Guarantees (ICML 2024 · similarity 0.87)
- Sketched Gaussian Mechanism for Private Federated Learning (NeurIPS 2025 · similarity 0.86)
- FedP3: Federated Personalized and Privacy-friendly Network Pruning under Model Heterogeneity (ICLR 2024 · similarity 0.86)
- A Bayesian Approach for Personalized Federated Learning in Heterogeneous Settings (NeurIPS 2024 · similarity 0.86)
- One Arrow, Two Hawks: Sharpness-aware Minimization for Federated Learning via Global Model Trajectory (ICML 2025 · similarity 0.86)
- Interaction-Aware Gaussian Weighting for Clustered Federated Learning (ICML 2025 · 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.
