AI Paper Atlas

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

FedLTN: Federated Learning for Sparse and Personalized Lottery Ticket Networks

Vaikkunth Mugunthan, Eric Lin, Vignesh Gokul, Christian Lau, Lalana Kagal, Steve Pieper

ECCV 2022 · Federated learning and privacy › Communication-efficient FL · approach: Communication-efficient heterogeneous FL · 7 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
postprune without rewinding to round 0 (ablation)FedLTN(0.9) · CIFAR-10average client test accuracy (%): 71.5 -> 82.6Improved
postprune without rewinding to round 0 (ablation)FedLTN(0.9) · Tiny ImageNetaverage client test accuracy (%): 71.8 -> 74.5Improved
postprune without rewinding to round 0 (ablation)FedLTN(0.9) · CIFAR-10minimum client test accuracy (%): 53.0 -> 63.0Improved
add Jump-Start local pretraining (ablation)FedLTN(0.9) · CIFAR-10communication cost (MB): 3,753.9 -> 1,846.0Improved
add Jump-Start local pretraining (ablation)FedLTN(0.9) · Tiny ImageNetaverage client test accuracy (%): 74.5 -> 83.1Improved
add Jump-Start local pretraining (ablation)FedLTN(0.9) · CIFAR-10average client test accuracy (%): 82.6 -> 82.2No clear effect
raise target pruning rate from 0.1 to 0.9 (ablation)FedLTN · CIFAR-10average client test accuracy (%): 74.5 -> 82.6Improved
raise target pruning rate from 0.1 to 0.9 (ablation)FedLTN · Tiny ImageNetaverage client test accuracy (%): 68.4 -> 74.5Improved
use FedLTN(0.5) instead of LotteryFL(0.5) with Dirichlet skewFedLTN · CIFAR-10 Dirichlet alpha=0.5average client test accuracy (%): 61.19 -> 67.20Improved
use FedLTN(0.9) instead of LotteryFL(0.5) with Dirichlet skewFedLTN · CIFAR-10 Dirichlet alpha=0.7average client test accuracy (%): 54.00 -> 55.60Mixed
use BN-preserved aggregation (ResNet18 vs custom CNN without BN)FedLTN · CIFAR-10average client test accuracy (%): 82.6 -> 75.4Improved
use FedLTN(0.9) instead of LotteryFL(0.5) in 100-client settingFedLTN · CIFAR-10 (100 clients, custom CNN)average client test accuracy (%): 74.9 -> 75.4Improved

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