AI Paper Atlas

Research map / Optimization / Distributed and compressed

On Biased Compression for Distributed Learning

Aleksandr Beznosikov, Samuel Horváth, Peter Richtárik, Mher Safaryan

JMLR 2023 · Optimization › Distributed and compressed · approach: Error feedback theory

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 error feedback to Top-5 compressed SGD (ablation)Top-5 SGD · VGG19 CIFAR10, 4 nodestraining/test loss and accuracyImproved
use Top-5 (biased) instead of Rand-5 (unbiased)compressed SGD · VGG19 CIFAR10training/test performanceImproved
add natural dithering after Top-kTop-k compression · VGG11 CIFAR10communication vs convergenceImproved
use natural compression instead of no compressionALBERT-large · GLUE MRPCscore: 82.5 -> 87.5Improved
use Power compression instead of no compressionALBERT-large · GLUE RTEscore: 66.3 -> 46.3Hurt
use Power compression instead of no compressionALBERT-large · communication round timeavg time per round: 9.62 -> 1.04Improved
use natural compression instead of no compressionALBERT-large · communication round timeavg time per round: 9.62 -> 4.05Improved
use Top-5 instead of Rand-5 sparsificationGaussian N(0;1) vectors · information savings d=100expected savings: 5 -> 27.14Improved

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