{"id":"arxiv/2610.01651","url":"https://arxiv.org/abs/2610.01651","tags":["math.PR","math.NA","stat.CO"],"score":0,"title":"Theoretical guarantees for stochastic gradient Langevin dynamics","domain":"math.PR","source":"arxiv","authors":["Daniel Paulin","Peter A. Whalley"],"reviews":0,"subject":"math.PR","summary":"We prove asymptotic bias bounds for stochastic gradient Langevin dynamics in Wasserstein distance of order two. We assume that the negative log-density is strongly convex with a Lipschitz gradient, and that the stochastic gradient estimator is unbiased with an error satisfying a mean-square Lipschitz condition. The bounds are of order $h$ under a fourth moment assumption on the stochastic gradient error and of order $h^{1/2}$ under only a second moment assumption, where $h$ is the stepsize. A spiked-noise example shows that a second moment assumption alone is insufficient for a bound of…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-01","reactions":{}}