Linear-Size Neural Network Representation of Piecewise Affine Functions in $\mathbb{R}^2$

Fuente: arXiv
Saved in:
Bibliographic Details
Main Author: Zanotti, Leo
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929762146975744
author Zanotti, Leo
author_facet Zanotti, Leo
contents It is shown that any continuous piecewise affine (CPA) function $\mathbb{R}^2\to\mathbb{R}$ with $p$ pieces can be represented by a ReLU neural network with two hidden layers and $O(p)$ neurons. Unlike prior work, which focused on convex pieces, this analysis considers CPA functions with connected but potentially non-convex pieces.
format Preprint
id arxiv_https___arxiv_org_abs_2503_13001
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Linear-Size Neural Network Representation of Piecewise Affine Functions in $\mathbb{R}^2$
Zanotti, Leo
Machine Learning
Neural and Evolutionary Computing
Metric Geometry
It is shown that any continuous piecewise affine (CPA) function $\mathbb{R}^2\to\mathbb{R}$ with $p$ pieces can be represented by a ReLU neural network with two hidden layers and $O(p)$ neurons. Unlike prior work, which focused on convex pieces, this analysis considers CPA functions with connected but potentially non-convex pieces.
title Linear-Size Neural Network Representation of Piecewise Affine Functions in $\mathbb{R}^2$
topic Machine Learning
Neural and Evolutionary Computing
Metric Geometry
url https://arxiv.org/abs/2503.13001