AI Paper Atlas

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

Breaking the centralized barrier for cross-device federated learning

Sai Praneeth Karimireddy, Martin Jaggi, Satyen Kale, Mehryar Mohri, Sashank J. Reddi, Sebastian U Stich, Ananda Theertha Suresh

NeurIPS 2021 · Federated learning and privacy › Communication-efficient FL · approach: Client participation analysis · 30 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
use Mime (SVRG correction + server optimizer state at each client step) instead of FedAvg, Adam baseAdam base optimizer · EMNIST CNNaccuracy (%): 85.9 -> 86.6Improved
use MimeAdam instead of FedAvgAdamAdam base optimizer · Shakespeareaccuracy (%): 18.5 -> 54.1Improved
use MimeAdam instead of FedAvgAdamAdam base optimizer · StackOverflowaccuracy (%): 3.2 -> 22.8Improved
use MimeSGD (SVRG correction) instead of FedAvgSGDSGD base optimizer · StackOverflowaccuracy (%): 23.8 -> 12.5Hurt
use MimeAdagrad instead of FedAvgAdagradAdagrad base optimizer · StackOverflowaccuracy (%): 24.2 -> 14.7Hurt
use MimeMom instead of FedAvgMommomentum base optimizer · Shakespeareaccuracy (%): 55.4 -> 53.6Hurt
use Mime-style local application of momentum instead of server-only momentum (ablation)Mime · EMNIST62 MLP, 10 local epochsaccuracyImproved
keep optimizer state fixed globally instead of updating it locally (Loc-Mime) (ablation)Mime · EMNIST62 MLP, 10 local epochsaccuracyImproved
drop the SVRG correction (MimeLite vs Mime) (ablation)Mime · EMNIST62 MLP, 10 local epochsaccuracyNo clear effect
use Mime/MimeLite instead of FedAvgSGD+momentum, MLP · EMNIST62 MLProunds to converge: about 7x fasterImproved
use Mime instead of SCAFFOLD/FedProxSGD+momentum · EMNIST62 MLPaccuracyImproved

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