Non-Uniform Diffusion Models

07/20/2022
by   Georgios Batzolis, et al.
0

Diffusion models have emerged as one of the most promising frameworks for deep generative modeling. In this work, we explore the potential of non-uniform diffusion models. We show that non-uniform diffusion leads to multi-scale diffusion models which have similar structure to this of multi-scale normalizing flows. We experimentally find that in the same or less training time, the multi-scale diffusion model achieves better FID score than the standard uniform diffusion model. More importantly, it generates samples 4.4 times faster in 128× 128 resolution. The speed-up is expected to be higher in higher resolutions where more scales are used. Moreover, we show that non-uniform diffusion leads to a novel estimator for the conditional score function which achieves on par performance with the state-of-the-art conditional denoising estimator. Our theoretical and experimental findings are accompanied by an open source library MSDiff which can facilitate further research of non-uniform diffusion models.

READ FULL TEXT

page 1

page 10

page 11

page 20

page 21

page 22

research
11/26/2021

Conditional Image Generation with Score-Based Diffusion Models

Score-based diffusion models have emerged as one of the most promising f...
research
08/06/2016

Weighted diffusion LMP algorithm for distributed estimation in non-uniform noise conditions

This letter presents an improved version of diffusion least mean ppower ...
research
07/01/2021

Benchmarking (multi)wavelet-based dynamic and static non-uniform grid solvers for flood inundation modelling

This paper explores static non-uniform grid solvers that adapt three ras...
research
02/10/2023

Example-Based Sampling with Diffusion Models

Much effort has been put into developing samplers with specific properti...
research
03/01/2023

Understanding the Diffusion Objective as a Weighted Integral of ELBOs

Diffusion models in the literature are optimized with various objectives...
research
08/19/2022

Learning filtered discretization operators: non-intrusive versus intrusive approaches

Simulating multi-scale phenomena such as turbulent fluid flows is typica...
research
12/16/2020

Testing the Stationarity Assumption in Software Effort Estimation Datasets

Software effort estimation (SEE) models are typically developed based on...

Please sign up or login with your details

Forgot password? Click here to reset