Skip to content

Official repository for "Nearest Neighbour Score Estimators for Diffusion Generative Models"

Notifications You must be signed in to change notification settings

plai-group/knn-score

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Nearest Neighbour Score Estimators for Diffusion Generative Models
Official PyTorch implementation of the ICML 2024 paper

Nearest Neighbour Score Estimators for Diffusion Generative Models
Matthew Niedoba, Dylan Green, Saeid Naderiparizi, Vasileios Lioutas, Jonathan Wilder Lavington, Xiaoxuan Liang, Yunpeng Liu, Ke Zhang, Setareh Dabiri, Adam Ścibior, Berend Zwartsenberg, Frank Wood
https://arxiv.org/abs/2402.08018

Abstract: Score function estimation is the cornerstone of both training and sampling from diffusion generative models. Despite this fact, the most commonly used estimators are either biased neural network approximations or high variance Monte Carlo estimators based on the conditional score. We introduce a novel nearest neighbour score function estimator which utilizes multiple samples from the training set to dramatically decrease estimator variance. We leverage our low variance estimator in two compelling applications. Training consistency models with our estimator, we report a significant increase in both convergence speed and sample quality. In diffusion models, we show that our estimator can replace a learned network for probability-flow ODE integration, opening promising new avenues of future research.

Overview

This repository is contains the code used to train consistency models using nearest neighbour score estimators. It is built upon the EDM repository. Environment details and code examples coming soon.

About

Official repository for "Nearest Neighbour Score Estimators for Diffusion Generative Models"

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages