English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Distributed asynchronous SGD has become widely used for deep learning inlarge-scale systems, but remains notorious for its instability when increasingthe number of workers. In this work, we study the dynamics of distributedasynchronous SGD under the lens of Lagrangian mechanics. Using thisdescription, we introduce the concept of energy to describe the optimizationprocess and derive a sufficient condition ensuring its stability as long as thecollective energy induced by the active workers remains below the energy of atarget synchronous process. Making use of this criterion, we derive a stabledistributed asynchronous optimization procedure, GEM, that estimates andmaintains the energy of the asynchronous system below or equal to the energy ofsequential SGD with momentum. Experimental results highlight the stability andspeedup of GEM compared to existing schemes, even when scaling to one hundredasynchronous workers. Results also indicate better generalization compared tothe targeted SGD with momentum.

Gradient Energy Matching for Distributed Asynchronous Gradient Descent