The Computational Complexity of Counting Linear Regions in ReLU Neural Networks

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Stargalla, Moritz, Hertrich, Christoph, Reichman, Daniel
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911361649344512
author Stargalla, Moritz
Hertrich, Christoph
Reichman, Daniel
author_facet Stargalla, Moritz
Hertrich, Christoph
Reichman, Daniel
contents An established measure of the expressive power of a given ReLU neural network is the number of linear regions into which it partitions the input space. There exist many different, non-equivalent definitions of what a linear region actually is. We systematically assess which papers use which definitions and discuss how they relate to each other. We then analyze the computational complexity of counting the number of such regions for the various definitions. Generally, this turns out to be an intractable problem. We prove NP- and #P-hardness results already for networks with one hidden layer and strong hardness of approximation results for two or more hidden layers. Finally, on the algorithmic side, we demonstrate that counting linear regions can at least be achieved in polynomial space for some common definitions.
format Preprint
id arxiv_https___arxiv_org_abs_2505_16716
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle The Computational Complexity of Counting Linear Regions in ReLU Neural Networks
Stargalla, Moritz
Hertrich, Christoph
Reichman, Daniel
Computational Complexity
Discrete Mathematics
Machine Learning
Neural and Evolutionary Computing
Combinatorics
An established measure of the expressive power of a given ReLU neural network is the number of linear regions into which it partitions the input space. There exist many different, non-equivalent definitions of what a linear region actually is. We systematically assess which papers use which definitions and discuss how they relate to each other. We then analyze the computational complexity of counting the number of such regions for the various definitions. Generally, this turns out to be an intractable problem. We prove NP- and #P-hardness results already for networks with one hidden layer and strong hardness of approximation results for two or more hidden layers. Finally, on the algorithmic side, we demonstrate that counting linear regions can at least be achieved in polynomial space for some common definitions.
title The Computational Complexity of Counting Linear Regions in ReLU Neural Networks
topic Computational Complexity
Discrete Mathematics
Machine Learning
Neural and Evolutionary Computing
Combinatorics
url https://arxiv.org/abs/2505.16716