{"id":"arxiv/2609.17750","url":"https://arxiv.org/abs/2609.17750","tags":["math.NA","math.PR","stat.CO"],"score":0,"title":"On the computational cost of Stochastic Gradient Langevin Dynamics","domain":"math.NA","source":"arxiv","authors":["Mateusz B. Majka","Tigran Nagapetyan","Łukasz Szpruch","Yue Wu","Danqi Zhuang"],"reviews":0,"subject":"math.NA","summary":"Stochastic Gradient Langevin Dynamics (SGLD) reduces the cost of Langevin-based sampling by replacing full-dataset drift evaluations with mini-batch approximations, but the resulting subsampling error may offset this computational saving. We study this trade-off for stochastic differential equations with finite-sum drifts and compare the computational cost of SGLD with that of the Euler-Maruyama (EM) method. For a prescribed mean-square accuracy $\\varepsilon^2$, we derive complexity estimates that explicitly track the dependence on the dataset size $m$, mini-batch size $s$, and accuracy…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-09-15","reactions":{}}