AI Paper Atlas

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

FedX: Unsupervised Federated Learning with Cross Knowledge Distillation

Sungwon Han, Sungwon Park, Fangzhao Wu, Sundong Kim, Chuhan Wu, Xing Xie, Meeyoung Cha

ECCV 2022 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Non-IID federated learning · 86 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 FedX two-sided knowledge distillationFedSimCLR · CIFAR-10 (last round)top-1 accuracy: 51.31 -> 56.88Improved
add FedXFedProtoCL · SVHN (last round)top-1 accuracy: 49.85 -> 69.31Improved
add FedXFedProtoCL · F-MNIST (best round)top-1 accuracy: 83.57 -> 83.34Hurt
add FedXFedMoCo · CIFAR-10 (last round)top-1 accuracy: 56.74 -> 58.23Improved
add global knowledge distillation to local KD (ablation)L_local-KD · CIFAR-10 (last round)top-1 accuracy: 51.89 -> 56.88Improved
replace global KD with SCAFFOLD or FedProx (ablation)L_local-KD + L_global-KD · CIFAR-10 (last round)top-1 accuracy: 56.88 -> 52.73 (SCAFFOLD), 52.48 (FedProx)Hurt
use global KD instead of FedProx (ablation)L_local-KD · SVHN (last round)top-1 accuracy: 77.43 -> 77.19No clear effect
add FedX across data sizesFedSimCLR · CIFAR-10 (10% data, last round)top-1 accuracy: 46.80 -> 51.03Improved
add FedX with 15 clientsFedSimCLR · CIFAR-10 (15 clients, last round)top-1 accuracy: 52.31 -> 55.12Improved
reduce data size from 100% to 10% (ablation)FedSimCLR + FedX · CIFAR-10 (last round)top-1 accuracy: 56.88 -> 51.03Hurt
add FedXFedSimCLR · ImageNet-10 (last round)top-1 accuracy: 81.50 -> 86.17Improved
use FedX instead of FedCA (ablation)FedSimCLR · CIFAR-10 (last round)top-1 accuracy: 47.46 -> 56.88Improved

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