{"id":"arxiv/2611.18738","url":"https://arxiv.org/abs/2611.18738","tags":["math.ST","math.PR"],"score":0,"title":"Support recovery of diffusion models under the manifold hypothesis","domain":"math.ST","source":"arxiv","authors":["Ayya Alieva","Gautam Iyer","Alexei Novikov","Lenya Ryzhik"],"reviews":0,"subject":"math.ST","summary":"We study diffusion models under the manifold hypothesis with possibly large score estimation and discretization errors, and show that while they may not exactly recover the data distribution, they will exactly recover the support (or data manifold). In continuous time, we use a Lyapunov function to show that even if the relative score estimation error is fixed (and not necessarily vanishing), then the rate at which the process approaches the support is $O(\\sqrt{(T-t)|\\log(T-t)|})$ as $t \\to T$, where $T$ is the final time. In discrete time, we use geometric time steps with a fixed…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-11-11","reactions":{}}