Saved in:
Bibliographic Details
Main Author: Hieu, Le Duc
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2509.08954
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909781723185152
author Hieu, Le Duc
author_facet Hieu, Le Duc
contents We study when the \emph{optimization curve} of first-order methods -- the sequence \${f(x\_n)}*{n\ge0}\$ produced by constant-stepsize iterations -- is convex, equivalently when the forward differences \$f(x\_n)-f(x*{n+1})\$ are nonincreasing. For gradient descent (GD) on convex \$L\$-smooth functions, the curve is convex for all stepsizes \$η\le 1.75/L\$, and this threshold is tight. Moreover, gradient norms are nonincreasing for all \$η\le 2/L\$, and in continuous time (gradient flow) the curve is always convex. These results complement and refine the classical smooth convex optimization toolbox, connecting discrete and continuous dynamics as well as worst-case analyses.
format Preprint
id arxiv_https___arxiv_org_abs_2509_08954
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Convexity of Optimization Curves: Local Sharp Thresholds, Robustness Impossibility, and New Counterexamples
Hieu, Le Duc
Optimization and Control
Machine Learning
90C25, 90C30, 65K05, 37N40, 26B25
We study when the \emph{optimization curve} of first-order methods -- the sequence \${f(x\_n)}*{n\ge0}\$ produced by constant-stepsize iterations -- is convex, equivalently when the forward differences \$f(x\_n)-f(x*{n+1})\$ are nonincreasing. For gradient descent (GD) on convex \$L\$-smooth functions, the curve is convex for all stepsizes \$η\le 1.75/L\$, and this threshold is tight. Moreover, gradient norms are nonincreasing for all \$η\le 2/L\$, and in continuous time (gradient flow) the curve is always convex. These results complement and refine the classical smooth convex optimization toolbox, connecting discrete and continuous dynamics as well as worst-case analyses.
title Convexity of Optimization Curves: Local Sharp Thresholds, Robustness Impossibility, and New Counterexamples
topic Optimization and Control
Machine Learning
90C25, 90C30, 65K05, 37N40, 26B25
url https://arxiv.org/abs/2509.08954