Denoising Diffusion Probabilistic Models
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.
Cite
Text
Ho et al. "Denoising Diffusion Probabilistic Models." Neural Information Processing Systems, 2020.Markdown
[Ho et al. "Denoising Diffusion Probabilistic Models." Neural Information Processing Systems, 2020.](https://mlanthology.org/neurips/2020/ho2020neurips-denoising/)BibTeX
@inproceedings{ho2020neurips-denoising,
title = {{Denoising Diffusion Probabilistic Models}},
author = {Ho, Jonathan and Jain, Ajay N. and Abbeel, Pieter},
booktitle = {Neural Information Processing Systems},
year = {2020},
url = {https://mlanthology.org/neurips/2020/ho2020neurips-denoising/}
}