Nearly Optimal Approximation Rates for Deep Super ReLU Networks on Sobolev Spaces

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Yang, Yahong, Wu, Yue, Yang, Haizhao, Xiang, Yang
Formato: Preprint
Publicado: 2023
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866909761354596352
author Yang, Yahong
Wu, Yue
Yang, Haizhao
Xiang, Yang
author_facet Yang, Yahong
Wu, Yue
Yang, Haizhao
Xiang, Yang
contents This paper introduces deep super ReLU networks (DSRNs) as a method for approximating functions in Sobolev spaces measured by Sobolev norms $W^{m,p}$ for $m\in\mathbb{N}$ with $m\ge 2$ and $1\le p\le +\infty$. Standard ReLU deep neural networks (ReLU DNNs) cannot achieve this goal. DSRNs consist primarily of ReLU DNNs, and several layers of the square of ReLU added at the end to smooth the networks output. This approach retains the advantages of ReLU DNNs, leading to the straightforward training. The paper also proves the optimality of DSRNs by estimating the VC-dimension of higher-order derivatives of DNNs, and obtains the generalization error in Sobolev spaces via an estimate of the pseudo-dimension of higher-order derivatives of DNNs.
format Preprint
id arxiv_https___arxiv_org_abs_2310_10766
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Nearly Optimal Approximation Rates for Deep Super ReLU Networks on Sobolev Spaces
Yang, Yahong
Wu, Yue
Yang, Haizhao
Xiang, Yang
Numerical Analysis
This paper introduces deep super ReLU networks (DSRNs) as a method for approximating functions in Sobolev spaces measured by Sobolev norms $W^{m,p}$ for $m\in\mathbb{N}$ with $m\ge 2$ and $1\le p\le +\infty$. Standard ReLU deep neural networks (ReLU DNNs) cannot achieve this goal. DSRNs consist primarily of ReLU DNNs, and several layers of the square of ReLU added at the end to smooth the networks output. This approach retains the advantages of ReLU DNNs, leading to the straightforward training. The paper also proves the optimality of DSRNs by estimating the VC-dimension of higher-order derivatives of DNNs, and obtains the generalization error in Sobolev spaces via an estimate of the pseudo-dimension of higher-order derivatives of DNNs.
title Nearly Optimal Approximation Rates for Deep Super ReLU Networks on Sobolev Spaces
topic Numerical Analysis
url https://arxiv.org/abs/2310.10766