Saved in:
Bibliographic Details
Main Authors: Zhang, Zihan, Shi, Lei, Zhou, Ding-Xuan
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.14899
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912722804801536
author Zhang, Zihan
Shi, Lei
Zhou, Ding-Xuan
author_facet Zhang, Zihan
Shi, Lei
Zhou, Ding-Xuan
contents In this paper, we study the binary classification problem on $[0,1]^d$ under the Tsybakov noise condition (with exponent $s \in [0,\infty]$) and the compositional assumption. This assumption requires the conditional class probability function of the data distribution to be the composition of $q+1$ vector-valued multivariate functions, where each component function is either a maximum value function or a Hölder-$β$ smooth function that depends only on $d_*$ of its input variables. Notably, $d_*$ can be significantly smaller than the input dimension $d$. We prove that, under these conditions, the optimal convergence rate for the excess 0-1 risk of classifiers is $\left( \frac{1}{n} \right)^{\frac{β\cdot(1\wedgeβ)^q}{{\frac{d_*}{s+1}+(1+\frac{1}{s+1})\cdotβ\cdot(1\wedgeβ)^q}}}$, which is independent of the input dimension $d$. Additionally, we demonstrate that ReLU deep neural networks (DNNs) trained with hinge loss can achieve this optimal convergence rate up to a logarithmic factor. This result provides theoretical justification for the excellent performance of ReLU DNNs in practical classification tasks, particularly in high-dimensional settings. The generalized approach is of independent interest.
format Preprint
id arxiv_https___arxiv_org_abs_2506_14899
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Optimal Convergence Rates of Deep Neural Network Classifiers
Zhang, Zihan
Shi, Lei
Zhou, Ding-Xuan
Machine Learning
In this paper, we study the binary classification problem on $[0,1]^d$ under the Tsybakov noise condition (with exponent $s \in [0,\infty]$) and the compositional assumption. This assumption requires the conditional class probability function of the data distribution to be the composition of $q+1$ vector-valued multivariate functions, where each component function is either a maximum value function or a Hölder-$β$ smooth function that depends only on $d_*$ of its input variables. Notably, $d_*$ can be significantly smaller than the input dimension $d$. We prove that, under these conditions, the optimal convergence rate for the excess 0-1 risk of classifiers is $\left( \frac{1}{n} \right)^{\frac{β\cdot(1\wedgeβ)^q}{{\frac{d_*}{s+1}+(1+\frac{1}{s+1})\cdotβ\cdot(1\wedgeβ)^q}}}$, which is independent of the input dimension $d$. Additionally, we demonstrate that ReLU deep neural networks (DNNs) trained with hinge loss can achieve this optimal convergence rate up to a logarithmic factor. This result provides theoretical justification for the excellent performance of ReLU DNNs in practical classification tasks, particularly in high-dimensional settings. The generalized approach is of independent interest.
title Optimal Convergence Rates of Deep Neural Network Classifiers
topic Machine Learning
url https://arxiv.org/abs/2506.14899