English

Memory and Communication Efficient Distributed Stochastic Optimization with Minibatch-Prox

Machine Learning 2017-06-12 v2

Abstract

We present and analyze an approach for distributed stochastic optimization which is statistically optimal and achieves near-linear speedups (up to logarithmic factors). Our approach allows a communication-memory tradeoff, with either logarithmic communication but linear memory, or polynomial communication and a corresponding polynomial reduction in required memory. This communication-memory tradeoff is achieved through minibatch-prox iterations (minibatch passive-aggressive updates), where a subproblem on a minibatch is solved at each iteration. We provide a novel analysis for such a minibatch-prox procedure which achieves the statistical optimal rate regardless of minibatch size and smoothness, thus significantly improving on prior work.

Keywords

Cite

@article{arxiv.1702.06269,
  title  = {Memory and Communication Efficient Distributed Stochastic Optimization with Minibatch-Prox},
  author = {Jialei Wang and Weiran Wang and Nathan Srebro},
  journal= {arXiv preprint arXiv:1702.06269},
  year   = {2017}
}
R2 v1 2026-06-22T18:23:48.029Z