Towards an Optimal Control Perspective of ResNet Training

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Püttschneider, Jens, Heilig, Simon, Fischer, Asja, Faulwasser, Timm
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911024661135360
author Püttschneider, Jens
Heilig, Simon
Fischer, Asja
Faulwasser, Timm
author_facet Püttschneider, Jens
Heilig, Simon
Fischer, Asja
Faulwasser, Timm
contents We propose a training formulation for ResNets reflecting an optimal control problem that is applicable for standard architectures and general loss functions. We suggest bridging both worlds via penalizing intermediate outputs of hidden states corresponding to stage cost terms in optimal control. For standard ResNets, we obtain intermediate outputs by propagating the state through the subsequent skip connections and the output layer. We demonstrate that our training dynamic biases the weights of the unnecessary deeper residual layers to vanish. This indicates the potential for a theory-grounded layer pruning strategy.
format Preprint
id arxiv_https___arxiv_org_abs_2506_21453
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Towards an Optimal Control Perspective of ResNet Training
Püttschneider, Jens
Heilig, Simon
Fischer, Asja
Faulwasser, Timm
Machine Learning
Systems and Control
Optimization and Control
We propose a training formulation for ResNets reflecting an optimal control problem that is applicable for standard architectures and general loss functions. We suggest bridging both worlds via penalizing intermediate outputs of hidden states corresponding to stage cost terms in optimal control. For standard ResNets, we obtain intermediate outputs by propagating the state through the subsequent skip connections and the output layer. We demonstrate that our training dynamic biases the weights of the unnecessary deeper residual layers to vanish. This indicates the potential for a theory-grounded layer pruning strategy.
title Towards an Optimal Control Perspective of ResNet Training
topic Machine Learning
Systems and Control
Optimization and Control
url https://arxiv.org/abs/2506.21453