AI Paper Atlas

Research map / Optimization / Distributed and compressed

Leader Stochastic Gradient Descent for Distributed Training of Deep Learning Models

Yunfei Teng, Wenbo Gao, François Chalus, Anna E Choromanska, Donald Goldfarb, Adrian Weller

NeurIPS 2019 · Optimization › Distributed and compressed · approach: Local and asynchronous SGD · 4 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
replace EASGD's averaged elastic force with a pull toward the leader worker (LSGD), 16 workersCNN7 distributed training · CIFAR-10convergence speed: roughly 1.5 times faster than EASGDImproved
LSGD instead of EASGD with 16 workersCNN7 distributed training · CIFAR-10test errorImproved
LSGD instead of EASGD with 4 workersVGG16 distributed training · CIFAR-10test errorNo clear effect
LSGD instead of DOWNPOURVGG16 distributed training · CIFAR-10convergence speed and test errorImproved
LSGD instead of EASGD with 4 workersResNet20 distributed training · CIFAR-10test errorImproved
LSGD instead of EASGD with 16 workersResNet20 distributed training · CIFAR-10convergence speed: roughly 2 times faster than EASGDImproved
LSGD and EASGD instead of DOWNPOUR and SGDCNN7 and ResNet20 on CIFAR-10 · CIFAR-10test errorImproved
LSGD instead of EASGD with 4 workersResNet50 distributed training · ImageNettest errorNo clear effect
LSGD instead of DOWNPOURResNet50 distributed training · ImageNetconvergence speedImproved

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