AI Paper Atlas

Research map / Federated learning and privacy / Personalized and heterogeneous FL

On Bridging Generic and Personalized Federated Learning for Image Classification

Hong-You Chen, Wei-Lun Chao

ICLR 2022 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Sparse and collaboration-graph personalization · 66 citations (OpenAlex)

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 changedWhereMeasuredEffect
Replace empirical risk with BRM in FedAvg (ablation)FedAvg · CIFAR-10 Dir(0.3)G-FL accuracy: 68.6 -> 76.8Improved
Replace empirical risk with BRM in FedAvg (ablation)FedAvg · CIFAR-10 Dir(0.3)P-FL accuracy: 85.1 -> 76.1Hurt
Add a personalized predictor trained with client empirical risk (ablation)FedAvg (BRM) · CIFAR-10 Dir(0.3)P-FL accuracy: 76.1 -> 86.4Improved
Replace the client-specific linear personalized head with a hypernetwork-generated head (ablation)Fed-RoD (linear, BRM) · CIFAR-10 Dir(0.3)P-FL accuracy: 86.4 -> 86.8Improved
Fine-tune the FedAvg (BRM) model on client data (ablation)FedAvg (BRM) · CIFAR-10 Dir(0.3)P-FL accuracy: 76.1 -> 84.5Improved
Use balanced softmax rather than cross entropy for BRM (ablation)Fed-RoD · CIFAR-10 Dir(0.3)G-FL accuracy: 68.6 -> 76.9Improved
Use balanced softmax rather than cross entropy for BRM (ablation)Fed-RoD · CIFAR-10 Dir(0.3)P-FL accuracy: 85.1 -> 86.8Improved
Combine FedDyn with Fed-RoDFed-RoD (hyper) · FMNIST Dir(0.1)G-FL accuracy: 83.9 -> 85.9Improved
Combine FedDyn with Fed-RoDFed-RoD (hyper) · FMNIST Dir(0.1)P-FL accuracy: 92.9 -> 95.3Improved
Combine FedDyn with Fed-RoDFed-RoD (hyper) · CIFAR-10 Dir(0.3)G-FL accuracy: 76.9 -> 74.6Hurt
Combine FedDyn with Fed-RoDFed-RoD (hyper) · CIFAR-10 Dir(0.3)P-FL accuracy: 86.8 -> 85.6Hurt
Use a hypernetwork conditioned on class distribution to produce personalized predictors for new clients (ablation)Fed-RoD (linear) · CIFAR-10 Dir(0.3), future clientsP-FL accuracy before fine-tuning: 62.4 -> 75.7Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy