Minibatch vs Local SGD with Shuffling: Tight Convergence Bounds and Beyond

Chulhee Yun · Shashank Rajput · Suvrit Sra

Keywords: [ convex optimization ] [ stochastic optimization ] [ distributed learning ] [ large scale learning ] [ local sgd ] [ federated learning ]

[ Abstract ]
[ Visit Poster at Spot B2 in Virtual World ] [ OpenReview
Wed 27 Apr 2:30 a.m. PDT — 4:30 a.m. PDT
Oral presentation: Oral 2: Understanding Deep Learning
Tue 26 Apr 1 a.m. PDT — 2:45 a.m. PDT


In distributed learning, local SGD (also known as federated averaging) and its simple baseline minibatch SGD are widely studied optimization methods. Most existing analyses of these methods assume independent and unbiased gradient estimates obtained via with-replacement sampling. In contrast, we study shuffling-based variants: minibatch and local Random Reshuffling, which draw stochastic gradients without replacement and are thus closer to practice. For smooth functions satisfying the Polyak-Ɓojasiewicz condition, we obtain convergence bounds (in the large epoch regime) which show that these shuffling-based variants converge faster than their with-replacement counterparts. Moreover, we prove matching lower bounds showing that our convergence analysis is tight. Finally, we propose an algorithmic modification called synchronized shuffling that leads to convergence rates faster than our lower bounds in near-homogeneous settings.

Chat is not available.