{"id":"arxiv/2610.19444","url":"https://arxiv.org/abs/2610.19444","tags":["math.OC","cs.LG","stat.ML"],"score":0,"title":"Large-Step Last-Iterate Convergence of Random-Reshuffling Stochastic Gradient Descent","domain":"math.OC","source":"arxiv","authors":["Zijun Chen","Zihan Zhang"],"reviews":0,"subject":"math.OC","summary":"Random reshuffling stochastic gradient descent (RR-SGD) is widely used for finite-sum optimization in machine learning. For convex $L$-smooth components, recent work establishes convergence guarantees for an averaged iterate with component stepsizes as large as order $1/L$. However, the final iterate is a natural output in practice, and existing last-iterate guarantees in this setting require stepsizes that scale inversely with the number of components $n$. We establish an expected objective-gap bound for the last iterate of standard RR-SGD on unconstrained problems with a constant component…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-14","reactions":{}}