Blurring Diffusion Models

arXiv:2209.05557v3 Announce Type: replace
Abstract: Recently, Rissanen et al., (2022) have presented a new type of diffusion process for generative modeling based on heat dissipation, or blurring, as an alternative to isotropic Gaussian diffusion. Here, we show that blurring can equivalently be defined through a Gaussian diffusion process with non-isotropic noise. In making this connection, we bridge the gap between inverse heat dissipation and denoising diffusion, and we shed light on the inductive bias that results from this modeling choice. …

abstract alternative arxiv bridge cs.cv cs.lg diffusion diffusion models generative generative modeling heat making modeling noise process show stat.ml through type