Id: arxiv/2611.11354 Url: https://arxiv.org/abs/2611.11354 Tags: math.OC Score: 0 Title: An Adaptive Short-Step Variant of BroxGD with Linear Convergence under Quadratic Growth Domain: math.OC Source: arxiv Authors: Jan Niklas Kolzenburg, Peter Richtárik Reviews: 0 Subject: math.OC Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-11-08 Page: https://delphi.science/r/arxiv/2611.11354 Broximal Gradient Descent (BroxGD), recently analyzed by Richtárik et al. (arXiv:2605.08850v1), is a projection-free sister method to Projected Gradient Descent (ProjGD) for solving $\min_{x\in\mathcal X}f(x)$, where $f:\mathbb R^d\to\mathbb R$ and $\mathcal X\subseteq\mathbb R^d$ is nonempty, closed and convex. The method performs local linear minimization over the feasible set intersected with a Euclidean ball and carefully prescribed radius. Like ProjGD, BroxGD enjoys linear convergence in the smooth strongly convex setting, with the same condition-number dependence. However, the known…