State space representations of the Roesser type for convolutional layers

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Pauli, Patricia, Gramlich, Dennis, Allgöwer, Frank
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917720541364224
author Pauli, Patricia
Gramlich, Dennis
Allgöwer, Frank
author_facet Pauli, Patricia
Gramlich, Dennis
Allgöwer, Frank
contents From the perspective of control theory, convolutional layers (of neural networks) are 2-D (or N-D) linear time-invariant dynamical systems. The usual representation of convolutional layers by the convolution kernel corresponds to the representation of a dynamical system by its impulse response. However, many analysis tools from control theory, e.g., involving linear matrix inequalities, require a state space representation. For this reason, we explicitly provide a state space representation of the Roesser type for 2-D convolutional layers with $c_\mathrm{in}r_1 + c_\mathrm{out}r_2$ states, where $c_\mathrm{in}$/$c_\mathrm{out}$ is the number of input/output channels of the layer and $r_1$/$r_2$ characterizes the width/length of the convolution kernel. This representation is shown to be minimal for $c_\mathrm{in} = c_\mathrm{out}$. We further construct state space representations for dilated, strided, and N-D convolutions.
format Preprint
id arxiv_https___arxiv_org_abs_2403_11938
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle State space representations of the Roesser type for convolutional layers
Pauli, Patricia
Gramlich, Dennis
Allgöwer, Frank
Systems and Control
Machine Learning
Image and Video Processing
Signal Processing
From the perspective of control theory, convolutional layers (of neural networks) are 2-D (or N-D) linear time-invariant dynamical systems. The usual representation of convolutional layers by the convolution kernel corresponds to the representation of a dynamical system by its impulse response. However, many analysis tools from control theory, e.g., involving linear matrix inequalities, require a state space representation. For this reason, we explicitly provide a state space representation of the Roesser type for 2-D convolutional layers with $c_\mathrm{in}r_1 + c_\mathrm{out}r_2$ states, where $c_\mathrm{in}$/$c_\mathrm{out}$ is the number of input/output channels of the layer and $r_1$/$r_2$ characterizes the width/length of the convolution kernel. This representation is shown to be minimal for $c_\mathrm{in} = c_\mathrm{out}$. We further construct state space representations for dilated, strided, and N-D convolutions.
title State space representations of the Roesser type for convolutional layers
topic Systems and Control
Machine Learning
Image and Video Processing
Signal Processing
url https://arxiv.org/abs/2403.11938