Locally orderless networks

Publikation: Bog/antologi/afhandling/rapport › Rapport

Standard

Locally orderless networks. / Sporring, Jon; Xu, Peidi; Lu, Jiahao; Lauze, Francois Bernard; Darkner, Sune.

arxiv.org, 2024. 12 s.

Publikation: Bog/antologi/afhandling/rapport › Rapport

Harvard

Sporring, J, Xu, P, Lu, J, Lauze, FB & Darkner, S 2024, Locally orderless networks. arxiv.org. <https://arxiv.org/abs/2406.13514>

APA

Sporring, J., Xu, P., Lu, J., Lauze, F. B., & Darkner, S. (2024). Locally orderless networks. arxiv.org. https://arxiv.org/abs/2406.13514

Vancouver

Sporring J, Xu P, Lu J, Lauze FB, Darkner S. Locally orderless networks. arxiv.org, 2024. 12 s.

Author

Sporring, Jon ; Xu, Peidi ; Lu, Jiahao ; Lauze, Francois Bernard ; Darkner, Sune. / Locally orderless networks. arxiv.org, 2024. 12 s.

Bibtex

@book{19fedb446546465ca663a25992ac272c,
title = "Locally orderless networks",
abstract = "We present Locally Orderless Networks (LON) and its theoretic foundation which links it to Convolutional Neural Networks (CNN), to Scale-space histograms, and measurement theory. The key elements are a regular sampling of the bias and the derivative of the activation function. We compare LON, CNN, and Scale-space histograms on prototypical single-layer networks. We show how LON and CNN can emulate each other, how LON expands the set of functionals computable to non-linear functions such as squaring. We demonstrate simple networks which illustrate the improved performance of LON over CNN on simple tasks for estimating the gradient magnitude squared, for regressing shape area and perimeter lengths, and for explainability of individual pixels' influence on the result. ",
author = "Jon Sporring and Peidi Xu and Jiahao Lu and Lauze, {Francois Bernard} and Sune Darkner",
year = "2024",
month = jun,
day = "19",
language = "English",
publisher = "arxiv.org",

}

RIS

TY - RPRT

T1 - Locally orderless networks

AU - Sporring, Jon

AU - Xu, Peidi

AU - Lu, Jiahao

AU - Lauze, Francois Bernard

AU - Darkner, Sune

PY - 2024/6/19

Y1 - 2024/6/19

N2 - We present Locally Orderless Networks (LON) and its theoretic foundation which links it to Convolutional Neural Networks (CNN), to Scale-space histograms, and measurement theory. The key elements are a regular sampling of the bias and the derivative of the activation function. We compare LON, CNN, and Scale-space histograms on prototypical single-layer networks. We show how LON and CNN can emulate each other, how LON expands the set of functionals computable to non-linear functions such as squaring. We demonstrate simple networks which illustrate the improved performance of LON over CNN on simple tasks for estimating the gradient magnitude squared, for regressing shape area and perimeter lengths, and for explainability of individual pixels' influence on the result.

AB - We present Locally Orderless Networks (LON) and its theoretic foundation which links it to Convolutional Neural Networks (CNN), to Scale-space histograms, and measurement theory. The key elements are a regular sampling of the bias and the derivative of the activation function. We compare LON, CNN, and Scale-space histograms on prototypical single-layer networks. We show how LON and CNN can emulate each other, how LON expands the set of functionals computable to non-linear functions such as squaring. We demonstrate simple networks which illustrate the improved performance of LON over CNN on simple tasks for estimating the gradient magnitude squared, for regressing shape area and perimeter lengths, and for explainability of individual pixels' influence on the result.

M3 - Report

BT - Locally orderless networks

PB - arxiv.org

ER -

ID: 395380287