Denoising Diffusion Probabilistic Models

TLDR

The paper demonstrates high‑quality image synthesis with diffusion probabilistic models, a latent variable framework inspired by nonequilibrium thermodynamics. The authors train models using a weighted variational bound linked to denoising score matching with Langevin dynamics, enabling a progressive lossy decompression scheme akin to autoregressive decoding. The models achieve an Inception score of 9.46 and FID of 3.17 on CIFAR‑10, and produce LSUN‑256×256 samples comparable to ProgressiveGAN. Code is available at https://github.com/hojonathanho/diffusion.

Abstract

We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at https://github.com/hojonathanho/diffusion