# Optimal Convergence Rates of Deep Neural Network Classifiers

arXiv: https://arxiv.org/abs/2506.14899 · HTML: https://mlsanity.com/paper/2506.14899/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 17 Jun 2025.

## Built on by (most cited first, 1 of 1)

- [Rate-optimal neural boundary detection from unlabeled noisy images](https://mlsanity.com/paper/2606.00715/) (2026)

## Abstract (arXiv)

In this paper, we study the binary classification problem on [0,1]d under the Tsybakov noise condition (with exponent s ∈ [0,∞]) 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 H"older-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( frac1n right)^fracβ·(1wedgeβ)qfracd_*s+1+(1+frac1s+1)·β·(1wedgeβ)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.