AI Paper Atlas

Research map / Federated learning and privacy / Communication-efficient FL

FedFit: Federated Dynamic Sparse Training via Fisher Information scoring

Meng Bi, Hong Huang, Jinlong Song, Charles Wang, Chengming Hu, Xi Chen, Ting Yu, Xue Liu

ICML 2026 · Federated learning and privacy › Communication-efficient FL · approach: Efficient local training in FL

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 Fisher-based scoring in the pruning stage (ablation)FedFit · CIFAR-10 (Fig. 6)mean accuracyImproved
add Fisher-based scoring in the growing stage (ablation)FedFit · CIFAR-10 (Fig. 6)mean accuracyImproved
use FedFit Fisher prune/grow scores instead of FedDST heuristics (ablation)FedDST · CIFAR-10 (Fig. 6)mean accuracyImproved
use FedFit instead of strongest baseline across datasetsstrongest baseline · Table 1 (CV tasks)test accuracy: roughly 3.8%Improved
use FedFit instead of FedRTS at low densityFedRTS · CIFAR10 ResNet18 rho=0.1test accuracy: 64.54 -> 65.51Improved
use FedFit instead of FedRTS at low densityFedRTS · SVHN ShuffleNetV2 rho=0.1test accuracy: 33.67 -> 50.63Improved
use FedFit instead of FedRTSFedRTS · Tiny ImageNet ResNet18 rho=0.2test accuracy: 5.60 -> 5.38Hurt
use FedFit instead of FedSGC at rho=0.3FedSGC · SVHN ShuffleNetV2 rho=0.3test accuracy: 70.24 -> 69.15Hurt
use FedFit at density 0.3 instead of dense FedAvgFedAvg · CIFAR10 ResNet18test accuracy: 77.25 -> 76.43Hurt
use FedFit at density 0.3 instead of dense FedAvgFedAvg · Tiny ImageNet ResNet18test accuracy: 8.97 -> 7.61Hurt
lower participation ratio of clientsFedFit · CIFAR-10 (Fig. 5)accuracyMixed
increase data heterogeneity (smaller Dirichlet a)FedFit · CIFAR-10 (Fig. 4)accuracyNo clear effect

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