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. Our implementation is available at this https URL
Cited in this thesis
Frequently Cited Together
- Minimally Invasive Evaluation of Venous Leg Ulcers in an Outpatient Setting Usin1 chapter
- Fish mislabelling in France: substitution rates and retail types1 chapter
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learni1 chapter
- Application of rapid evaporative ionization mass spectrometry in preclinical and1 chapter
- Qualitative and quantitative analysis of adulterated Antarctic Krill Oil (AKO) b1 chapter
- DNA barcoding reveals mislabeling of endangered sharks sold as swordfish in New 1 chapter
BibTeX
@article{Ho2020,
author = {Ho, Jonathan and Jain, Ajay and Abbeel, Pieter},
journal = {Advances in Neural Information Processing Systems},
title = {Denoising diffusion probabilistic models},
year = {2020},
pages = {6840–6851},
volume = {33},
}