{"id":"arxiv/2610.37444","url":"https://arxiv.org/abs/2610.37444","tags":["math.OC"],"score":0,"title":"DREAM: Delay-Resistant Acceleration with Momentum","domain":"math.OC","source":"arxiv","authors":["Mert Gürbüzbalaban","M. Hamza Rafi","Necdet Serhat Aybat","Waheed U. Bajwa"],"reviews":0,"subject":"math.OC","summary":"With exact, delay-free gradients, Nesterov's accelerated gradient (NAG) achieves optimal $\\sqrtκ$ dependence for $L$-smooth, $μ$-strongly convex objectives, where $κ=L/μ$; heavy-ball (HB) achieves the same scaling on strongly convex quadratics. In asynchronous and distributed settings, gradients can be inexact, with delays varying across workers and iterations. We isolate staleness using a fixed delay $τ\\ge1$, replacing $\\nabla f(y_k)$ with $\\nabla f(y_{k-τ})$ at iteration $k$. Arjevani, Shamir, and Srebro establish a worst-case lower bound of $Ω(τ\\sqrtκ\\log(1/\\varepsilon))$ iterations to…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-27","reactions":{}}