Research map / Federated learning and privacy / Personalized and heterogeneous FL
A Bayesian Federated Learning Framework With Online Laplace Approximation
Liangxi Liu, Xi Jiang, Feng Zheng, Hong Chen, Guo-Jun Qi, Heng Huang, Ling Shao
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 |
|---|---|---|---|
| use multivariate Gaussian product aggregation with prior-loss local training instead of FedAvg | FedAvg · CIFAR-10 CNN, alpha=0.01, N=20 | global accuracy: 42.16% -> 48.57% | Improved |
| use FOLA instead of FedAvg for convergence speed | FedAvg · CIFAR-10 CNN, alpha=0.01, N=20 | rounds to reach 30% GA: 23 -> 4 rounds to 30% GA | Improved |
| use FOLA instead of FedAvg for convergence speed to 40% GA | FedAvg · CIFAR-10 CNN, alpha=0.01, N=20 | rounds to reach 40% GA: 73 -> 21 rounds | Improved |
| use FedCurv instead of FedAvg (indirect AE reduction) | FedAvg · CIFAR-10 CNN, alpha=0.01, N=20 | global accuracy: 42.16% -> 41.31% | No clear effect |
| use FedNova under extreme heterogeneity | FedAvg · CIFAR-10 CNN, alpha=0.01, N=20 | global accuracy: 42.16% -> 10.00% | Hurt |
| use SCAFFOLD under extreme heterogeneity | FedAvg · CIFAR-10 CNN, alpha=0.01, N=20 | global accuracy: 42.16% -> 20.30% | Hurt |
| use online Laplace approximation instead of offline (ablation) | FOLA aggregation · CIFAR-10 CNN | global accuracy | Improved |
| add Gaussian-product aggregation over FedAvg averaging (same local training) (ablation) | FOLA · CIFAR-10 CNN | global accuracy | Improved |
| add both aggregation and local prior-loss strategies versus single module (ablation) | FOLA · CIFAR-10 CNN | global accuracy | Improved |
| use full Laplace instead of diagonal Fisher covariance (ablation) | FOLA aggregation · MNIST 780-10, 2 clients | global accuracy: nearly 90% GA in 3 rounds vs more than 10 rounds | Improved |
| use e-full instead of diagonal Fisher covariance (ablation) | FOLA aggregation · MNIST 780-10, 2 clients | global accuracy | Hurt |
| decrease data heterogeneity alpha (more heterogeneous) (ablation) | FOLA vs FedAvg · MNIST MLP | global accuracy | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- FedBE — FedBE: Making Bayesian Model Ensemble Applicable to Federated Learning (ICLR 2021) · better on 4 of 4 shared results
- SCAFFOLD — SCAFFOLD: Stochastic Controlled Averaging for Federated Learning (ICML 2020) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Improving Generalization in Federated Learning with Model-Data Mutual Information Regularization: A Posterior Inference Approach (NeurIPS 2024 · similarity 0.89)
- Calibrated One Round Federated Learning with Bayesian Inference in the Predictive Space (AAAI 2024 · similarity 0.89)
- Personalized Bayesian Federated Learning with Wasserstein Barycenter Aggregation (NeurIPS 2025 · similarity 0.88)
- A Bayesian Approach for Personalized Federated Learning in Heterogeneous Settings (NeurIPS 2024 · similarity 0.88)
- Harnessing Heterogeneous Statistical Strength for Personalized Federated Learning via Hierarchical Bayesian Inference (ICML 2025 · similarity 0.87)
- FedBEns: One-Shot Federated Learning based on Bayesian Ensemble (ICML 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.
