SciPost Submission Page

How to GAN Event Unweighting

by Mathias Backes, Anja Butter, Tilman Plehn, Ramon Winterhalder

Submission summary

As Contributors: Tilman Plehn · Ramon Winterhalder
Arxiv Link: https://arxiv.org/abs/2012.07873v3 (pdf)
Date accepted: 2021-04-21
Date submitted: 2021-03-25 10:44
Submitted by: Winterhalder, Ramon
Submitted to: SciPost Physics
Academic field: Physics
Specialties:
  • High-Energy Physics - Phenomenology
Approach: Computational

Abstract

Event generation with neural networks has seen significant progress recently. The big open question is still how such new methods will accelerate LHC simulations to the level required by upcoming LHC runs. We target a known bottleneck of standard simulations and show how their unweighting procedure can be improved by generative networks. This can, potentially, lead to a very significant gain in simulation speed.

Published as SciPost Phys. 10, 089 (2021)




Reports on this Submission

Anonymous Report 2 on 2021-4-7 Invited Report

  • Cite as: Anonymous, Report on arXiv:2012.07873v3, delivered 2021-04-07, doi: 10.21468/SciPost.Report.2762

Report

Dear authors,

thanks for clarifying the jargon in the new version.

Did you have any thoughts on my previous question as to where that small horizontal shift between the uwGAN curve and true distribution in the bottom left/right plots of Fig1 might come from (also visible as a slope in the bottom left plot of Fig3)?

Requested changes

No specific changes requested.

  • validity: -
  • significance: -
  • originality: -
  • clarity: -
  • formatting: -
  • grammar: -

Author Ramon Winterhalder on 2021-04-12
(in reply to Report 2 on 2021-04-07)

Did you have any thoughts on my previous question as to where that small horizontal shift between the uwGAN curve and true distribution in the bottom left/right plots of Fig1 might come from (also visible as a slope in the bottom left plot of Fig3)?

-> The mentioned small shift is not reproducible over various training runs. Minor deviations between the uwGAN
curve and the true distribution are always visible and are an effect of limited training size and limited statistics.

Anonymous Report 1 on 2021-3-29 Invited Report

Report

The authors have addressed my comments sufficiently. I am happy for publication to proceed.

  • validity: -
  • significance: -
  • originality: -
  • clarity: -
  • formatting: -
  • grammar: -

Login to report or comment