AI Paper Atlas

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

Submodel Extraction for Efficient and Personalized Federated Learning via Optimal Transport

Zheng Jiang, Nan He, Yiming Chen, Lifeng Sun

CVPR 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
replace OT-enhanced aggregation with position-based aggregation (i.e. add OTA) (ablation)SubFLOT · Digit5accuracy (%): 78.74 -> 92.58Improved
add OTA aggregation (ablation)SubFLOT · CIFAR10 beta=0.1accuracy (%): 81.41 -> 83.78Improved
add SAR regularization (ablation)SubFLOT · Digit5accuracy (%): 89.73 -> 92.58Improved
add SAR regularization (ablation)SubFLOT · CIFAR10 beta=0.1accuracy (%): 82.77 -> 83.78Improved
use OT-based pruning instead of fixed-position pruning (ablation)SubFLOT · Digit5accuracy (%): 90.37 -> 92.58Improved
use OT-based pruning instead of fixed-position pruning (ablation)SubFLOT · CIFAR10 beta=0.1accuracy (%): 83.21 -> 83.78Improved
use historical client model as OTP proxy instead of random pruning (ablation)SubFLOT OTP · CIFAR10 beta=0.1accuracy (%): 49.82 -> 83.78Improved
use historical client model as OTP proxy instead of magnitude pruning (ablation)SubFLOT OTP · CIFAR10 beta=0.1accuracy (%): 77.04 -> 83.78Improved
increase fusion alpha from 0.5 to 0.7 (ablation)SubFLOT · CIFAR10 beta=0.1accuracy (%): 83.78 -> 84.39Improved
increase SAR lambda from 1 to 10 (ablation)SubFLOT · CIFAR10 beta=0.1accuracy (%): 83.78 -> 82.93Hurt
increase SAR lambda from 1 to 5 (ablation)SubFLOT · CIFAR10 beta=0.1accuracy (%): 83.78 -> 83.91No clear effect
use SubFLOT submodels instead of full-model FedAvgSubFLOT · VGG11 per-client communicationComm. (MB): 9.23 -> 4.33Improved

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