Towards Faster Non-Asymptotic Convergence for Diffusion-Based Generative Models

arXiv:2306.09251v3 Announce Type: replace-cross
Abstract: Diffusion models, which convert noise into new data instances by learning to reverse a Markov diffusion process, have become a cornerstone in contemporary generative modeling. While their practical power has now been widely recognized, the theoretical underpinnings remain far from mature. In this work, we develop a suite of non-asymptotic theory towards understanding the data generation process of diffusion models in discrete time, assuming access to $\ell_2$-accurate estimates of the (Stein) score functions. For a …

abstract arxiv become convergence cs.it cs.lg data diffusion diffusion models faster generative generative modeling generative models instances markov math.it math.st modeling noise power practical process stat.ml stat.th type work