On Statistical Rates of Conditional Diffusion Transformers: Approximation, Estimation and Minimax Optimality

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Hu, Jerry Yao-Chieh, Wu, Weimin, Lee, Yi-Chen, Huang, Yu-Chao, Chen, Minshuo, Liu, Han
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913586320769024
author Hu, Jerry Yao-Chieh
Wu, Weimin
Lee, Yi-Chen
Huang, Yu-Chao
Chen, Minshuo
Liu, Han
author_facet Hu, Jerry Yao-Chieh
Wu, Weimin
Lee, Yi-Chen
Huang, Yu-Chao
Chen, Minshuo
Liu, Han
contents We investigate the approximation and estimation rates of conditional diffusion transformers (DiTs) with classifier-free guidance. We present a comprehensive analysis for ``in-context'' conditional DiTs under four common data assumptions. We show that both conditional DiTs and their latent variants lead to the minimax optimality of unconditional DiTs under identified settings. Specifically, we discretize the input domains into infinitesimal grids and then perform a term-by-term Taylor expansion on the conditional diffusion score function under Hölder smooth data assumption. This enables fine-grained use of transformers' universal approximation through a more detailed piecewise constant approximation and hence obtains tighter bounds. Additionally, we extend our analysis to the latent setting under the linear latent subspace assumption. We not only show that latent conditional DiTs achieve lower bounds than conditional DiTs both in approximation and estimation, but also show the minimax optimality of latent unconditional DiTs. Our findings establish statistical limits for conditional and unconditional DiTs, and offer practical guidance toward developing more efficient and accurate DiT models.
format Preprint
id arxiv_https___arxiv_org_abs_2411_17522
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle On Statistical Rates of Conditional Diffusion Transformers: Approximation, Estimation and Minimax Optimality
Hu, Jerry Yao-Chieh
Wu, Weimin
Lee, Yi-Chen
Huang, Yu-Chao
Chen, Minshuo
Liu, Han
Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
We investigate the approximation and estimation rates of conditional diffusion transformers (DiTs) with classifier-free guidance. We present a comprehensive analysis for ``in-context'' conditional DiTs under four common data assumptions. We show that both conditional DiTs and their latent variants lead to the minimax optimality of unconditional DiTs under identified settings. Specifically, we discretize the input domains into infinitesimal grids and then perform a term-by-term Taylor expansion on the conditional diffusion score function under Hölder smooth data assumption. This enables fine-grained use of transformers' universal approximation through a more detailed piecewise constant approximation and hence obtains tighter bounds. Additionally, we extend our analysis to the latent setting under the linear latent subspace assumption. We not only show that latent conditional DiTs achieve lower bounds than conditional DiTs both in approximation and estimation, but also show the minimax optimality of latent unconditional DiTs. Our findings establish statistical limits for conditional and unconditional DiTs, and offer practical guidance toward developing more efficient and accurate DiT models.
title On Statistical Rates of Conditional Diffusion Transformers: Approximation, Estimation and Minimax Optimality
topic Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2411.17522