AI Paper Atlas

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

FedGain: Toward Negative-Gain-Free Client Collaboration in Federated Learning

Yuqing Zhang, Changli Zhou, Binghuang Huang, Hui Tian

ICML 2026 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Non-IID federated learning

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 FedGain collaboration structure to FedAvgFedAvg · FMNIST (Label Shift)accuracy: 46.62 -> 90.16Improved
add FedGain to FedAvgFedAvg · CIFAR10 (Feature Shift)accuracy: 44.74 -> 50.44Improved
add FedGain to FedAvgFedAvg · CIFAR100 (Quality Shift)accuracy: 26.27 -> 37.95Improved
add FedGain to personalized FL (pFedMe)pFedMe · CIFAR10 (Feature Shift)accuracy: 38.00 -> 47.48Improved
add FedGain to FedAS and FedCSD on FMNISTFedAS, FedCSD · FMNIST (Label Shift)NGP: FedAS 70.00 -> 5.00; FedCSD 65.00 -> 17.50Mixed
add FedGain to GFL and PFL methods on OfficeHomeFedAvg, FedProx, Ditto, etc. · OfficeHome (Domain Shift)NGP: FedAvg 25.00 -> 0.00Improved
use FedGain clustering instead of IFCA, FedCluster, FeSEM, CFL-GPclustered FL with FedAvg · CIFAR100 (Quality Shift)accuracy: 29.78 (FeSEM, best baseline) -> 37.95Improved
use FedGain clustering vs FedCluster on FMNISTclustered FL with FedAvg · FMNIST (Label Shift)accuracy: 90.11 (FedCluster) -> 90.16No clear effect
use only distribution shift term (remove quantity skew) (ablation)FedGain · FMNIST (Label Shift)accuracy: 52.26 -> 90.16 when adding quantity termImproved
use only quantity skew term (remove distribution shift) (ablation)FedGain · CIFAR100 (Quality Shift)accuracy: 34.44 -> 37.95 when adding distribution termImproved
set kernel decay gamma to 5.0 vs 2.0 or 8.0 (ablation)FedGain · CIFAR100 (Quality Shift)accuracy: 31.64 (2.0), 37.95 (5.0), 36.95 (8.0)Mixed
vary capacity constant C in {6, 8, 10} (ablation)FedGain · CIFAR10/FMNIST/CIFAR100accuracy: CIFAR100: 34.44, 37.95, 34.49Mixed

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