English

Random Reshuffling Dominates Stochastic Gradient Descent

Optimization and Control 2026-06-30 v1 Machine Learning Machine Learning

Abstract

Stochastic Gradient Descent (SGD\textsf{SGD}) is one of the most classical optimization algorithms with favorable theoretical guarantees, yet the practical implementation of SGD\textsf{SGD} differs subtly from its well-known form and is often referred to as Shuffling Stochastic Gradient Descent (Shuffling SGD\textsf{Shuffling SGD}). A particularly popular strategy in Shuffling SGD\textsf{Shuffling SGD} is Random Reshuffling (RR\textsf{RR}), which has achieved great empirical success across numerous experiments. Despite its strong performance, RR\textsf{RR} has long been considered a heuristic due to a lack of theoretical support. Over the last decade, people have finally established provable convergence rates for RR\textsf{RR}, thus justifying its observed superiority. However, for smooth convex optimization, two clouds over the convergence theory of RR\textsf{RR} remain to this day. More precisely, according to the current theory, Shuffling SGD\textsf{Shuffling SGD} under RR\textsf{RR} converges only when the stepsize is smaller than a threshold proportional to 1/n1/n, where nn is the number of summands in the objective (or the number of data points). Consequently, the optimally tuned theoretical rate of Shuffling SGD\textsf{Shuffling SGD} under RR\textsf{RR} is strictly worse than that of SGD\textsf{SGD} when the number of epochs is smaller than another threshold proportional to nn. These two restrictions heavily limit the applicability of existing theories and leave a critical mismatch with practice. In this work, for the first time, we prove that RR\textsf{RR} dominates SGD\textsf{SGD} in smooth convex optimization under any reasonable stepsize after any finite number of epochs, thereby addressing a longstanding open question.

Cite

@article{arxiv.2606.32005,
  title  = {Random Reshuffling Dominates Stochastic Gradient Descent},
  author = {Zijian Liu},
  journal= {arXiv preprint arXiv:2606.32005},
  year   = {2026}
}

Comments

COLT 2026