Cover Image

How to understand limitations of generative networks

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, th...

Full description

Saved in:
Bibliographic Details
Main Authors: Das, Ranit (Author) , Favaro, Luigi (Author) , Heimel, Theo (Author) , Krause, Claudius (Author) , Plehn, Tilman (Author) , Shih, David (Author)
Format: Article (Journal)
Language:English
Published: 25 January 2024
In: SciPost physics
Year: 2024, Volume: 16, Issue: 1, Pages: 1-32
ISSN:2542-4653
DOI:10.21468/SciPostPhys.16.1.031
Online Access:Verlag, lizenzpflichtig, Volltext: https://doi.org/10.21468/SciPostPhys.16.1.031
Verlag, lizenzpflichtig, Volltext: https://scipost.org/10.21468/SciPostPhys.16.1.031
Get full text
Author Notes:Ranit Das, Luigi Favaro, Theo Heimel, Claudius Krause, Tilman Plehn and David Shih
Description
Summary: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.
Item Description:Veröffentlicht: 25. Januar 2024
Gesehen am 05.06.2024
Physical Description:Online Resource
ISSN:2542-4653
DOI:10.21468/SciPostPhys.16.1.031

Similar Items