Research map / Federated learning and privacy / Personalized and heterogeneous FL
FedHB: Hierarchical Bayesian Federated Learning
Minyoung Kim, Timothy Hospedales
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 |
|---|---|---|---|
| model client-wise latent variables with a hierarchical Bayesian model (NIW) instead of FedAvg-style shared parameter | FedAvg · CIFAR-100 global prediction (s=100,f=0.1,tau=1) | test accuracy (%): 40.87 -> 49.76 | Improved |
| hierarchical Bayesian modelling (NIW) instead of pFedBayes shared theta | pFedBayes · CIFAR-100 global prediction (s=100,f=0.1,tau=1) | test accuracy (%): 37.23 -> 49.76 | Improved |
| use FedHB Mixture (K=2) instead of Fed-BABU | Fed-BABU · CIFAR-100 personalisation (s=100,f=0.1,tau=1) | test accuracy (%): 50.43 -> 56.17 | Improved |
| use FedHB NIW instead of Fed-BABU | Fed-BABU · CIFAR-100 personalisation (s=10,f=1.0,tau=1) | test accuracy (%): 78.92 -> 82.71 | Improved |
| use FedHB NIW instead of Fed-BABU | Fed-BABU · CIFAR-100 global prediction (s=100,f=0.1,tau=10) | test accuracy (%): 27.93 -> 29.02 | Improved |
| use FedHB Mixture (K=2) instead of Fed-BABU | Fed-BABU · CIFAR-C-100 personalisation, held-out corruptions (s=100,f=0.1,tau=1) | test accuracy (%): 69.93 -> 74.16 | Improved |
| use FedHB NIW instead of Fed-BABU | Fed-BABU · CIFAR-C-100 global prediction (s=100,f=0.1,tau=1) | test accuracy (%): 79.45 -> 81.22 | Improved |
| use FedHB Mixture (K=2) instead of Fed-BABU | Fed-BABU · MNIST global prediction (s=5,f=0.1,tau=1) | test accuracy (%): 97.16 -> 97.94 | Improved |
| use FedHB Mixture (K=2) instead of Fed-BABU | Fed-BABU · EMNIST personalisation (s=5,f=0.2,tau=1) | test accuracy (%): 83.09 -> 88.97 | Improved |
| Bayesian personalisation training after global training | FedHB · CIFAR-100 (s=100,f=0.1,tau=10) | test accuracy (%): about 4% | Improved |
| increase number of participating clients (ablation) | FedHB · MNIST training | convergence rate | Improved |
| use FedHB NIW instead of Fed-BABU | Fed-BABU · CIFAR-100 global prediction (s=10,f=1.0,tau=10) | test accuracy (%): 21.92 -> 22.87 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Fed-BABU — FedBABU: Toward Enhanced Representation for Federated Image Classification (ICLR 2022) · better on 13 of 13 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
- A Bayesian Approach for Personalized Federated Learning in Heterogeneous Settings (NeurIPS 2024 · similarity 0.90)
- Harnessing Heterogeneous Statistical Strength for Personalized Federated Learning via Hierarchical Bayesian Inference (ICML 2025 · similarity 0.89)
- Calibrated One Round Federated Learning with Bayesian Inference in the Predictive Space (AAAI 2024 · similarity 0.89)
- Improving Generalization in Federated Learning with Model-Data Mutual Information Regularization: A Posterior Inference Approach (NeurIPS 2024 · similarity 0.88)
- Personalized Bayesian Federated Learning with Wasserstein Barycenter Aggregation (NeurIPS 2025 · similarity 0.88)
- Federated Generalised Variational Inference: A Robust Probabilistic Federated Learning Framework (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.
