In recent years, variance-reducing stochastic methods have shown great practical performance, exhibiting linear convergence rate when other stochastic methods offered a sub-linear rate. However, as datasets grow ever bigger and clusters become widespread, the need for fast distribution methods is pressing. We propose here a distribution scheme for SAGA which maintains a linear convergence rate, even when communication between nodes is limited.
@article{arxiv.1705.10405,
title = {Distributed SAGA: Maintaining linear convergence rate with limited communication},
author = {Clément Calauzènes and Nicolas Le Roux},
journal= {arXiv preprint arXiv:1705.10405},
year = {2017}
}