Local and global topological complexity measures OF ReLU neural network functions

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Grigsby, J. Elisenda, Lindsey, Kathryn, Masden, Marissa
Format: Preprint
Published: 2022
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916188037054464
author Grigsby, J. Elisenda
Lindsey, Kathryn
Masden, Marissa
author_facet Grigsby, J. Elisenda
Lindsey, Kathryn
Masden, Marissa
contents We apply a generalized piecewise-linear (PL) version of Morse theory due to Grunert-Kuhnel-Rote to define and study new local and global notions of topological complexity for fully-connected feedforward ReLU neural network functions, F: R^n -> R. Along the way, we show how to construct, for each such F, a canonical polytopal complex K(F) and a deformation retract of the domain onto K(F), yielding a convenient compact model for performing calculations. We also give a construction showing that local complexity can be arbitrarily high.
format Preprint
id arxiv_https___arxiv_org_abs_2204_06062
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle Local and global topological complexity measures OF ReLU neural network functions
Grigsby, J. Elisenda
Lindsey, Kathryn
Masden, Marissa
Algebraic Topology
Computational Geometry
Machine Learning
Geometric Topology
57R70, 57Q99, 52B70, 52C35
We apply a generalized piecewise-linear (PL) version of Morse theory due to Grunert-Kuhnel-Rote to define and study new local and global notions of topological complexity for fully-connected feedforward ReLU neural network functions, F: R^n -> R. Along the way, we show how to construct, for each such F, a canonical polytopal complex K(F) and a deformation retract of the domain onto K(F), yielding a convenient compact model for performing calculations. We also give a construction showing that local complexity can be arbitrarily high.
title Local and global topological complexity measures OF ReLU neural network functions
topic Algebraic Topology
Computational Geometry
Machine Learning
Geometric Topology
57R70, 57Q99, 52B70, 52C35
url https://arxiv.org/abs/2204.06062