{"id":"arxiv/2601.18425","url":"https://arxiv.org/abs/2601.18425","tags":["math.NA"],"score":0,"title":"Stochstic Sampling for Generative Diffusion Models: From Euler-Maruyama to Higher-Order Schemes","domain":"math.NA","source":"arxiv","authors":["Emanuel Pfarr","Radu Timofte","Frank Werner"],"reviews":0,"subject":"math.NA","summary":"We develop a convergence analysis for generative diffusion models that simultaneously accounts for the three principal sources of error in stochastic sampling: initialization error, score-matching error, and discretization of the reverse-time SDE. Our central tool is the notion of a general strong scheme, a broad class of discretization methods for the reverse dynamics defined via explicit, index-wise tolerances on their Itô-Taylor coefficients. This notion extends the classical strong-scheme framework of Kloeden and Platen to an iterate-wise formulation, which is strictly stronger and…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-01-26","reactions":{}}