How to GAN away Detector Effects

TL;DR

This paper demonstrates how generative networks can invert detector simulations in LHC analyses, enabling reconstruction of parton-level data from measured events, thus improving the interpretability of experimental results.

Contribution

It introduces a method using fully conditional generative networks to invert Monte Carlo simulations of detector effects, a novel approach in high-energy physics analysis.

Findings

Successfully reconstructs parton-level information from measured events.

Shows how to train generative networks to invert complex simulations.

Introduces a technique to stagger or cool maximum mean discrepancy loss.

Abstract

LHC analyses directly comparing data and simulated events bear the danger of using first-principle predictions only as a black-box part of event simulation. We show how simulations, for instance, of detector effects can instead be inverted using generative networks. This allows us to reconstruct parton level information from measured events. Our results illustrate how, in general, fully conditional generative networks can statistically invert Monte Carlo simulations. As a technical by-product we show how a maximum mean discrepancy loss can be staggered or cooled.