{"id":"arxiv/2610.36205","url":"https://arxiv.org/abs/2610.36205","tags":["stat.ML","cs.LG","math.ST"],"score":0,"title":"Approximation Families and the Statistical Efficiency of Denoising Score Matching","domain":"stat.ML","source":"arxiv","authors":["Brian Miyatake"],"reviews":0,"subject":"stat.ML","summary":"We study score matching over finite-dimensional approximation spaces generated by gradients of eigenfunctions of the infinitesimal generator of a forward diffusion. We compare polynomial approximation under one-dimensional Ornstein-Uhlenbeck flow with trigonometric-polynomial approximation under heat flow on the torus. In the Ornstein-Uhlenbeck setting, we show that for every non-Gaussian initial distribution uniformly comparable to the standard Gaussian, the asymptotic score-estimation variance grows exponentially along a subsequence of polynomial degrees. Averaging the score-matching loss…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-10-26","reactions":{}}