{"id":"arxiv/2604.10728","url":"https://arxiv.org/abs/2604.10728","tags":["math.OC"],"score":0,"title":"Last Iterate Convergence of AdaGrad-Norm for Convex Non-Smooth Optimization","domain":"math.OC","source":"arxiv","authors":["Margarita Preobrazhenskaia","Makar Sidorov","Igor Preobrazhenskii","Eduard Gorbunov"],"reviews":0,"subject":"math.OC","summary":"We study the convergence of the last iterate (i.e., the $(N+1)$-th iterate) of the AdaGrad method. Although AdaGrad -- an adaptive subgradient method -- underpins a wide class of algorithms, most existing convergence analyses focus on averaged (or best) iterates. We derive worst-case upper bounds on the suboptimality of the final point and show that, with an optimally tuned stepsize parameter, the last iterate converges at the rate $O(1/N^{1/4})$. We complement this guarantee with matching lower-bound constructions, proving that this rate is tight and that AdaGrad's last-iterate rate is…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-04-12","reactions":{}}