AI Paper Atlas

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

Federated Learning with Unlabeled Clients: Personalization Can Happen in Low Dimensions

Hossein Zakerinia, Jonathan Scott, Christoph H. Lampert

ICML 2026 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Bayesian personalized FL · 0 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
add unlabeled clients to training (ablation)FLowDUP (CNN) · CIFAR10 p=0.1accuracy: 65.3 -> 67.5Improved
add unlabeled clients to training (ablation)FLowDUP (CNN) · CIFAR10 p=0.5accuracy: 82.5 -> 83.8No clear effect
add unlabeled clients to training (ablation)FLowDUP (ResNet18) · CIFAR10 p=0.2accuracy: 83.3 -> 83.0No clear effect
add unlabeled clients to training (ablation)FLowDUP (ResNet18) · CIFAR10 p=0.05accuracy: 54.8 -> 57.5Improved
learn the regularizer psi_r instead of fixing it to 0 (ablation)FLowDUP (CNN) · CIFAR10 p=1.0accuracy: 85.6 -> 86.6Improved
learn the regularizer psi_r instead of fixing it to 0 (ablation)FLowDUP (CNN) · CIFAR10 p=0.1accuracy: 65.5 -> 67.5Improved
increase subspace dimension k from 200 to 10000 (ablation)FLowDUP (CNN) · CIFAR10 p=1.0accuracy: 71.2 -> 86.6Improved
increase subspace dimension k from 200 to 10000 (ablation)FLowDUP (CNN) · CIFAR10 p=0.1accuracy: 56.6 -> 66.1Improved
use ResNet18 instead of CNN as client model f (h1 = CNN) (ablation)FLowDUP · CIFAR10 p=0.1accuracy: 66.1 -> 71.4Improved
use ResNet18 instead of CNN as hypernetwork encoder h1 (f = CNN) (ablation)FLowDUP · CIFAR10 p=0.5accuracy: 83.3 -> 87.3Improved
personalize with FLowDUP instead of FedTTAFLowDUP vs FedTTA · FEMNIST p=1.0accuracy: 74.3 -> 89.3Improved
use FLowDUP instead of FedAvgFLowDUP vs FedAvg · CIFAR10 CNN p=0.1accuracy: 47.9 -> 66.1Improved

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