Research map / Federated learning and privacy / Personalized and heterogeneous FL
Personalized Federated Learning With Gaussian Processes
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 |
|---|---|---|---|
| use shared deep kernel with personal GP (pFedGP) vs best baseline pFedHN | pFedGP · CIFAR-100, 500 clients | test accuracy (%): 34.1 -> 50.6 | Helped |
| pFedGP vs pFedHN with 100 clients | pFedGP · CIFAR-100, 100 clients | test accuracy (%): 52.3 -> 61.3 | Helped |
| pFedGP vs pFedHN with sufficient data (50 clients) | pFedGP · CIFAR-10, 50 clients | test accuracy (%): 90.2 -> 89.2 | No clear effect |
| add global inducing points for data efficiency (IP-data) vs full GP (ablation) | pFedGP · CIFAR-100, 500 clients | test accuracy (%): 50.6 -> 55.7 | Helped |
| use IP-data vs full GP with many samples (ablation) | pFedGP · CIFAR-100, 50 clients | test accuracy (%): 63.3 -> 60.2 | Hurt |
| use inducing points (IP-compute) instead of full GP (ablation) | pFedGP · CINIC-10, 50 clients | test accuracy (%): 71.3 -> 71.5 | No clear effect |
| use inducing points (IP-compute) for runtime, 400 IPs (ablation) | pFedGP · CINIC-10, 50 clients | inference run time (sec): 1.08 -> 0.42 | Helped |
| reduce IP-compute inducing points from 400 to 80 (ablation) | pFedGP-IP-compute · CINIC-10, 50 clients | test accuracy (%): 71.5 -> 70.2 | Hurt |
| pFedGP vs pFedHN under per-client input noise | pFedGP · Noisy CIFAR-100, 100 clients | test accuracy (%): 38.9 -> 49.5 | Helped |
| global inducing points under client-specific input noise (ablation) | pFedGP · Noisy CIFAR-100, 100 clients | test accuracy (%): 49.5 -> 45.0 | Hurt |
| fewer samples per client (800 to 80) for Local training | Local · CIFAR-100 | test accuracy (%): 52.1 -> 30.9 | Hurt |
Closest papers by meaning
- Understanding the Statistical Accuracy-Communication Trade-off in Personalized Federated Learning with Minimax Guarantees (ICML 2025 · similarity 0.90)
- FedAS: Bridging Inconsistency in Personalized Federated Learning (CVPR 2024 · similarity 0.90)
- PeFLL: Personalized Federated Learning by Learning to Learn (ICLR 2024 · similarity 0.90)
- pFedClub: Controllable Heterogeneous Model Aggregation for Personalized Federated Learning (NeurIPS 2024 · similarity 0.89)
- Harnessing Heterogeneous Statistical Strength for Personalized Federated Learning via Hierarchical Bayesian Inference (ICML 2025 · similarity 0.89)
- Decentralized Directed Collaboration for Personalized Federated Learning (CVPR 2024 · similarity 0.89)
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.