Research map / Federated learning and privacy / Personalized and heterogeneous FL
Personalized Federated Learning via Variational Bayesian Inference
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 variational Bayesian personalized models with a global distribution as prior | pFedBayes · MNIST, FMNIST, and CIFAR-10; small, medium, and large data | personalized accuracy (%): 1.25\%, 1.78\% and 0.52\% on MNIST; 0.42\%, 0.63\% and 0.79\% on FMNIST; 11.71\%, 7.19\% and 6.33\% on CIFAR-10 | Helped |
| use pFedBayes global distribution aggregation | pFedBayes · MNIST small, medium, and large | global accuracy (%): 2.79\%, 1.67\% and 1.97\% | Helped |
| use pFedBayes global distribution aggregation | pFedBayes · FMNIST small, medium, and large | global accuracy (%) | Hurt |
| use pFedBayes global distribution aggregation | pFedBayes · CIFAR-10 small, medium, and large | global accuracy (%): 3.47\% and 3.49\% on small and medium datasets | Mixed |
| set personalized and global learning rates both to 0.001 rather than 0.005 (ablation) | pFedBayes · MNIST medium | personalized accuracy (%): 97.18 -> 97.28 | Helped |
| set personalized and global learning rates both to 0.001 rather than 0.005 (ablation) | pFedBayes · MNIST medium | global accuracy (%): 93.22 -> 90.22 | Helped |
| set KL tradeoff zeta to 10 rather than 0.5, 1, 5, or 20 (ablation) | pFedBayes · MNIST medium | personalized accuracy (%): 97.13, 97.41, 97.38, 97.18, 97.04 | Mixed |
| set KL tradeoff zeta to 10 rather than 0.5, 1, 5, or 20 (ablation) | pFedBayes · MNIST medium | global accuracy (%): 89.88, 91.24, 93.03, 93.22, 92.77 | Helped |
| initialize rho to -2.5 rather than -1, -1.5, -2, or -3 (ablation) | pFedBayes · MNIST medium | personalized accuracy (%): 97.38, 97.38, 97.45, 97.18, 97.08 | Mixed |
| initialize rho to -2.5 rather than -1, -1.5, -2, or -3 (ablation) | pFedBayes · MNIST medium | global accuracy (%): 92.31, 92.40, 92.40, 93.22, 92.16 | Helped |
| increase pFedMe learning rates to 0.1 from 0.01 (ablation) | pFedMe · MNIST medium | personalized accuracy (%): 95.16 -> 19.73 | Hurt |
| increase HeurFedAMP learning rate to 0.1 from 0.01 at alpha 0.5 (ablation) | HeurFedAMP · MNIST medium | personalized accuracy (%): 94.95 -> 93.29 | Hurt |
Closest papers by meaning
- Harnessing Heterogeneous Statistical Strength for Personalized Federated Learning via Hierarchical Bayesian Inference (ICML 2025 · similarity 0.91)
- Understanding the Statistical Accuracy-Communication Trade-off in Personalized Federated Learning with Minimax Guarantees (ICML 2025 · similarity 0.89)
- Bayesian Coreset Optimization for Personalized Federated Learning (ICLR 2024 · similarity 0.89)
- Federated Generalised Variational Inference: A Robust Probabilistic Federated Learning Framework (ICML 2025 · similarity 0.89)
- A Bayesian Approach for Personalized Federated Learning in Heterogeneous Settings (NeurIPS 2024 · similarity 0.88)
- Personalized Bayesian Federated Learning with Wasserstein Barycenter Aggregation (NeurIPS 2025 · similarity 0.88)
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.