AI Paper Atlas

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

FedOpenMatch: Towards Semi-Supervised Federated Learning in Open-Set Environments

Hongquan Liu, ChenyuGuo, Yixin Ren, Jihong Guan, Shuigeng Zhou

ICLR 2026 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Federated semi-supervised and graph 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
use FedOpenMatch instead of federated IOMatchopen-set SSFL · CIFAR-100 80 seen/25 labels, Dir(0.1)closed-set accuracy: 43.28 -> 50.39Improved
use FedOpenMatch instead of federated IOMatchopen-set SSFL · CIFAR-100 80 seen/25 labels, Dir(0.1)open-set accuracy: 33.02 -> 47.35Improved
use FedOpenMatch on CIFAR10/SVHN with milder heterogeneityopen-set SSFL · CIFAR10 and SVHN, Dir(0.3)closed/open-set accuracy: 72.32 -> 73.32 (CIFAR10 closed, vs IOMatch)No clear effect
adapt centralized OSSL (OpenMatch, SSB) to federated learningOSSFL · CIFAR-100closed-set accuracy: e.g. SSB 5.14 vs labeled-only 17.29Hurt
add gradient stop between inlier and OVA branches (ablation)FedOpenMatch · CIFAR-100 80/25 Dir(0.1)open-set accuracy: 34.81 -> 39.51Improved
add logit adjustment for OVA imbalance (ablation)FedOpenMatch · CIFAR-100 80/25 Dir(0.1)open-set accuracy: 39.51 -> 43.65Improved
add logit consistency regularization (ablation)FedOpenMatch · CIFAR-100 80/25 Dir(0.1)open-set accuracy: 43.65 -> 48.36Improved
use logit consistency instead of probability consistency loss (ablation)FedOpenMatch · CIFAR-100 80/25 Dir(0.1)closed/open-set accuracy: +4.83% closed, +3.68% openImproved
raise consistency weight lambda from 1 to 2 (ablation)FedOpenMatch · CIFAR-100accuracy: closed +0.9%, open -4.7%Mixed

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