Saved in:
Bibliographic Details
Main Authors: Kuehn, Christian, Kuntz, Sara-Viola, Wöhrer, Tobias
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.28591
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915899903049728
author Kuehn, Christian
Kuntz, Sara-Viola
Wöhrer, Tobias
author_facet Kuehn, Christian
Kuntz, Sara-Viola
Wöhrer, Tobias
contents We analyze the universal approximation constraints of narrow Residual Neural Networks (ResNets) both theoretically and numerically. For deep neural networks without input space augmentation, a central constraint is the inability to represent critical points of the input-output map. We prove that this has global consequences for target function approximations and show that the manifestation of this defect is typically a shift of the critical point to infinity, which we call the ``tunnel effect'' in the context of classification tasks. While ResNets offer greater expressivity than standard multilayer perceptrons (MLPs), their capability strongly depends on the signal ratio between the skip and residual channels. We establish quantitative approximation bounds for both the residual-dominant (close to MLP) and skip-dominant (close to neural ODE) regimes. These estimates depend explicitly on the channel ratio and uniform network weight bounds. Low-dimensional examples further provide a detailed analysis of the different ResNet regimes and how architecture-target incompatibility influences the approximation error.
format Preprint
id arxiv_https___arxiv_org_abs_2603_28591
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Universal Approximation Constraints of Narrow ResNets: The Tunnel Effect
Kuehn, Christian
Kuntz, Sara-Viola
Wöhrer, Tobias
Dynamical Systems
Machine Learning
41A30, 58K05, 68T07
We analyze the universal approximation constraints of narrow Residual Neural Networks (ResNets) both theoretically and numerically. For deep neural networks without input space augmentation, a central constraint is the inability to represent critical points of the input-output map. We prove that this has global consequences for target function approximations and show that the manifestation of this defect is typically a shift of the critical point to infinity, which we call the ``tunnel effect'' in the context of classification tasks. While ResNets offer greater expressivity than standard multilayer perceptrons (MLPs), their capability strongly depends on the signal ratio between the skip and residual channels. We establish quantitative approximation bounds for both the residual-dominant (close to MLP) and skip-dominant (close to neural ODE) regimes. These estimates depend explicitly on the channel ratio and uniform network weight bounds. Low-dimensional examples further provide a detailed analysis of the different ResNet regimes and how architecture-target incompatibility influences the approximation error.
title Universal Approximation Constraints of Narrow ResNets: The Tunnel Effect
topic Dynamical Systems
Machine Learning
41A30, 58K05, 68T07
url https://arxiv.org/abs/2603.28591