Adversarial learning to eliminate systematic errors: a case study in High Energy Physics

Victor Estrade 1, 2, 3 Cécile Germain 2, 1, 3 Isabelle Guyon 1, 2, 3, 4 David Rousseau 5
1 TAU - TAckling the Underspecified
LRI - Laboratoire de Recherche en Informatique, Inria Saclay - Ile de France, CNRS - Centre National de la Recherche Scientifique : UMR8623, UP11 - Université Paris-Sud - Paris 11
Abstract : Making the region selection procedure used in High Energy Physics analysis robust to systematic errors is a case of supervised domain adaptation. This paper proposes a benchmark that captures a simple but realistic case of systematic HEP analysis, in order to expose the issue to the wider community. The benchmark makes easy to conduct an experimental comparison of the recent adversarial knowledge-free approach and a less data-intensive alternative.
Complete list of metadatas

Cited literature [13 references]  Display  Hide  Download

https://hal.inria.fr/hal-01665925
Contributor : Cecile Germain <>
Submitted on : Sunday, December 17, 2017 - 6:00:31 PM
Last modification on : Thursday, February 7, 2019 - 2:37:44 PM

File

FINALnipswkshversioncourte.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01665925, version 1

Citation

Victor Estrade, Cécile Germain, Isabelle Guyon, David Rousseau. Adversarial learning to eliminate systematic errors: a case study in High Energy Physics. NIPS 2017 - workshop Deep Learning for Physical Sciences, Dec 2017, Long Beach, United States. pp.1-5. ⟨hal-01665925⟩

Share

Metrics

Record views

479

Files downloads

218