Understanding Generalizability of Diffusion Models Requires Rethinking the Hidden Gaussian Structure

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Xiang, Dai, Yixiang, Qu, Qing
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916502186229760
author Li, Xiang
Dai, Yixiang
Qu, Qing
author_facet Li, Xiang
Dai, Yixiang
Qu, Qing
contents In this work, we study the generalizability of diffusion models by looking into the hidden properties of the learned score functions, which are essentially a series of deep denoisers trained on various noise levels. We observe that as diffusion models transition from memorization to generalization, their corresponding nonlinear diffusion denoisers exhibit increasing linearity. This discovery leads us to investigate the linear counterparts of the nonlinear diffusion models, which are a series of linear models trained to match the function mappings of the nonlinear diffusion denoisers. Surprisingly, these linear denoisers are approximately the optimal denoisers for a multivariate Gaussian distribution characterized by the empirical mean and covariance of the training dataset. This finding implies that diffusion models have the inductive bias towards capturing and utilizing the Gaussian structure (covariance information) of the training dataset for data generation. We empirically demonstrate that this inductive bias is a unique property of diffusion models in the generalization regime, which becomes increasingly evident when the model's capacity is relatively small compared to the training dataset size. In the case that the model is highly overparameterized, this inductive bias emerges during the initial training phases before the model fully memorizes its training data. Our study provides crucial insights into understanding the notable strong generalization phenomenon recently observed in real-world diffusion models.
format Preprint
id arxiv_https___arxiv_org_abs_2410_24060
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Understanding Generalizability of Diffusion Models Requires Rethinking the Hidden Gaussian Structure
Li, Xiang
Dai, Yixiang
Qu, Qing
Machine Learning
Computer Vision and Pattern Recognition
Image and Video Processing
Signal Processing
In this work, we study the generalizability of diffusion models by looking into the hidden properties of the learned score functions, which are essentially a series of deep denoisers trained on various noise levels. We observe that as diffusion models transition from memorization to generalization, their corresponding nonlinear diffusion denoisers exhibit increasing linearity. This discovery leads us to investigate the linear counterparts of the nonlinear diffusion models, which are a series of linear models trained to match the function mappings of the nonlinear diffusion denoisers. Surprisingly, these linear denoisers are approximately the optimal denoisers for a multivariate Gaussian distribution characterized by the empirical mean and covariance of the training dataset. This finding implies that diffusion models have the inductive bias towards capturing and utilizing the Gaussian structure (covariance information) of the training dataset for data generation. We empirically demonstrate that this inductive bias is a unique property of diffusion models in the generalization regime, which becomes increasingly evident when the model's capacity is relatively small compared to the training dataset size. In the case that the model is highly overparameterized, this inductive bias emerges during the initial training phases before the model fully memorizes its training data. Our study provides crucial insights into understanding the notable strong generalization phenomenon recently observed in real-world diffusion models.
title Understanding Generalizability of Diffusion Models Requires Rethinking the Hidden Gaussian Structure
topic Machine Learning
Computer Vision and Pattern Recognition
Image and Video Processing
Signal Processing
url https://arxiv.org/abs/2410.24060