{"id":"arxiv/2607.15980","url":"https://arxiv.org/abs/2607.15980","tags":["math.OC"],"score":0,"title":"Sharp Dimension Dependence for the Last Iterate of the SubGradient Method","domain":"math.OC","source":"arxiv","authors":["Guglielmo Beretta","Tommaso Cesari","Roberto Colomboni","Andrea Paudice"],"reviews":0,"subject":"math.OC","summary":"We study the last iterate of the projected subGradient Method (sGM) for convex Lipschitz objectives defined on $\\mathbb{R}^d$. We prove that, for a finite horizon $n$ and a constant stepsize $η=Θ(1/\\sqrt n)$, the last iterate achieves an optimization error of order $d/\\sqrt n$, showing that the extra $\\log n$ factor appearing in high dimensions is unnecessary in every fixed dimension. We complement this result with a matching linear-in-$d$ lower bound and show that the sharp worst-case dimension-horizon dependence is of order $\\min\\{d,\\log n\\}/\\sqrt n$. This solves, in particular, a COLT…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-07-17","reactions":{}}