How to understand limitations of generative networks

Das, Ranit; Favaro, Luigi; Heimel, Theo; Krause, Claudius; Plehn, Tilman; Shih, David

doi:10.21468/SciPostPhys.16.1.031

SciPost Physics

How to understand limitations of generative networks

Ranit Das, Luigi Favaro, Theo Heimel, Claudius Krause, Tilman Plehn, David Shih

SciPost Phys. 16, 031 (2024) · published 25 January 2024

doi: 10.21468/SciPostPhys.16.1.031
pdf
Submissions/Reports

Abstract

Well-trained classifiers and their complete weight distributions provide us with a well-motivated and practicable method to test generative networks in particle physics. We illustrate their benefits for distribution-shifted jets, calorimeter showers, and reconstruction-level events. In all cases, the classifier weights make for a powerful test of the generative network, identify potential problems in the density estimation, relate them to the underlying physics, and tie in with a comprehensive precision and uncertainty treatment for generative networks.

TY  - JOUR
PB  - SciPost Foundation
DO  - 10.21468/SciPostPhys.16.1.031
TI  - How to understand limitations of generative networks
PY  - 2024/01/25
UR  - https://scipost.org/SciPostPhys.16.1.031
JF  - SciPost Physics
JA  - SciPost Phys.
VL  - 16
IS  - 1
SP  - 031
A1  - Das, Ranit
AU  - Favaro, Luigi
AU  - Heimel, Theo
AU  - Krause, Claudius
AU  - Plehn, Tilman
AU  - Shih, David
AB  - Well-trained classifiers and their complete weight distributions provide us with a well-motivated and practicable method to test generative networks in particle physics. We illustrate their benefits for distribution-shifted jets, calorimeter showers, and reconstruction-level events. In all cases, the classifier weights make for a powerful test of the generative network, identify potential problems in the density estimation, relate them to the underlying physics, and tie in with a comprehensive precision and uncertainty treatment for generative networks.
ER  -

@Article{10.21468/SciPostPhys.16.1.031,
	title={{How to understand limitations of generative networks}},
	author={Ranit Das and Luigi Favaro and Theo Heimel and Claudius Krause and Tilman Plehn and David Shih},
	journal={SciPost Phys.},
	volume={16},
	pages={031},
	year={2024},
	publisher={SciPost},
	doi={10.21468/SciPostPhys.16.1.031},
	url={https://scipost.org/10.21468/SciPostPhys.16.1.031},
}

Cited by 9

Authors / Affiliations: mappings to Contributors and Organizations

See all Organizations.

¹ Ranit Das,
² Luigi Favaro,
² Theo Heimel,
² Claudius Krause,
² Tilman Plehn,
¹ David Shih

Funders for the research work leading to this publication