English

Accelerated Gradient Descent via Long Steps

Optimization and Control 2023-09-28 v2

Abstract

Recently Grimmer [1] showed for smooth convex optimization by utilizing longer steps periodically, gradient descent's textbook LD2/2TLD^2/2T convergence guarantees can be improved by constant factors, conjecturing an accelerated rate strictly faster than O(1/T)O(1/T) could be possible. Here we prove such a big-O gain, establishing gradient descent's first accelerated convergence rate in this setting. Namely, we prove a O(1/T1.0564)O(1/T^{1.0564}) rate for smooth convex minimization by utilizing a nonconstant nonperiodic sequence of increasingly large stepsizes. It remains open if one can achieve the O(1/T1.178)O(1/T^{1.178}) rate conjectured by Das Gupta et. al. [2] or the optimal gradient method rate of O(1/T2)O(1/T^2). Big-O convergence rate accelerations from long steps follow from our theory for strongly convex optimization, similar to but somewhat weaker than those concurrently developed by Altschuler and Parrilo [3].

Keywords

Cite

@article{arxiv.2309.09961,
  title  = {Accelerated Gradient Descent via Long Steps},
  author = {Benjamin Grimmer and Kevin Shu and Alex L. Wang},
  journal= {arXiv preprint arXiv:2309.09961},
  year   = {2023}
}
R2 v1 2026-06-28T12:25:07.901Z