Saved in:
Bibliographic Details
Main Authors: Grillo, Moritz, Hofmann, Tobias
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.14068
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909849618481152
author Grillo, Moritz
Hofmann, Tobias
author_facet Grillo, Moritz
Hofmann, Tobias
contents We study the expressivity of sparse maxout networks, where each neuron takes a fixed number of inputs from the previous layer and employs a, possibly multi-argument, maxout activation. This setting captures key characteristics of convolutional or graph neural networks. We establish a duality between functions computable by such networks and a class of virtual polytopes, linking their geometry to questions of network expressivity. In particular, we derive a tight bound on the dimension of the associated polytopes, which serves as the central tool for our analysis. Building on this, we construct a sequence of depth hierarchies. While sufficiently deep sparse maxout networks are universal, we prove that if the required depth is not reached, width alone cannot compensate for the sparsity of a fixed indegree constraint.
format Preprint
id arxiv_https___arxiv_org_abs_2510_14068
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle On the expressivity of sparse maxout networks
Grillo, Moritz
Hofmann, Tobias
Machine Learning
Artificial Intelligence
Combinatorics
68T07, 52B05, 14T99
We study the expressivity of sparse maxout networks, where each neuron takes a fixed number of inputs from the previous layer and employs a, possibly multi-argument, maxout activation. This setting captures key characteristics of convolutional or graph neural networks. We establish a duality between functions computable by such networks and a class of virtual polytopes, linking their geometry to questions of network expressivity. In particular, we derive a tight bound on the dimension of the associated polytopes, which serves as the central tool for our analysis. Building on this, we construct a sequence of depth hierarchies. While sufficiently deep sparse maxout networks are universal, we prove that if the required depth is not reached, width alone cannot compensate for the sparsity of a fixed indegree constraint.
title On the expressivity of sparse maxout networks
topic Machine Learning
Artificial Intelligence
Combinatorics
68T07, 52B05, 14T99
url https://arxiv.org/abs/2510.14068