Nearest Neighbour Score Estimators for Diffusion Generative Models

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 nov...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2024-07
Main Authors: Niedoba, Matthew, Green, Dylan, Naderiparizi, Saeid, Lioutas, Vasileios, Jonathan Wilder Lavington, Liang, Xiaoxuan, Liu, Yunpeng, Zhang, Ke, Dabiri, Setareh, Ścibior, Adam, Zwartsenberg, Berend, Wood, Frank
Format: Article
Language:English
Subjects:
Diffusion rate
Estimators
Neural networks
Training
Variance
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary: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.
ISSN:2331-8422