AI Paper Atlas

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

Few-Round Learning for Federated Learning

Younghyun Park, Dong-Jun Han, Do-Yeon Kim, Jun Seo, Jaekyun Moon

NeurIPS 2021 · Federated learning and privacy › Communication-efficient FL · approach: Efficient local training in 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
add global prototype-assisted learning (GPAL) to distance-based FRL (ablation)FRL · CIFAR-100 unseen, IIDaccuracy: 69.74 -> 72.93Improved
add GPAL to distance-based FRL (ablation)FRL · CIFAR-100 unseen, non-IIDaccuracy: 63.85 -> 66.87Improved
add GPAL to distance-based FRL (ablation)FRL · FEMNIST unseen, non-IIDaccuracy: 88.60 -> 92.42Improved
add GPAL to distance-based FRL (ablation)FRL · miniImageNet unseen, IIDaccuracy: 68.05 -> 69.31Improved
distance-based instead of linear classifier (ablation)FRL · CIFAR-100 unseen, non-IIDaccuracy: 52.98 -> 63.85Improved
distance-based instead of linear classifier (ablation)FRL · CIFAR-100 unseen, IIDaccuracy: 67.32 -> 69.74Improved
distance-based instead of linear classifier (ablation)FRL · miniImageNet unseen, IIDaccuracy: 67.75 -> 68.05Improved
meta-learned initialization (FRL linear) instead of FedAvg fine-tuning (ablation)FRL · miniImageNet unseen, non-IIDaccuracy: 33.39 -> 53.51Improved
pretrain with FedAvg before deployment instead of random init (ablation)FedAvg · CIFAR-100 unseen, IIDaccuracy: 51.55 -> 63.18Improved
pretrain with FedAvg before deployment instead of random init (ablation)FedAvg · FEMNIST unseen, non-IIDaccuracy: 59.22 -> 58.23No clear effect
use one-shot FL fine-tuning instead of FedAvg fine-tuning (ablation)fine-tuning baseline · CIFAR-100 unseen, non-IIDaccuracy: 44.33 -> 35.11Hurt
use one-shot FL fine-tuning instead of FedAvg fine-tuning (ablation)fine-tuning baseline · CIFAR-100 unseen, IIDaccuracy: 63.18 -> 64.71Improved

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