Invertible Networks or Partons to Detector and Back Again

TL;DR

This paper introduces invertible neural networks for particle physics simulations, enabling bidirectional mapping between detector data and underlying physics, with applications to ZW production and QCD jet unfolding at the LHC.

Contribution

It presents a novel application of conditional invertible neural networks to invert detector effects and QCD radiation, providing per-event probabilistic interpretations in high-energy physics.

Findings

Successful inversion of detector simulations for ZW production

Unfolding of QCD radiation to parton-level phase space

Per-event probabilistic interpretation achieved

Abstract

For simulations where the forward and the inverse directions have a physics meaning, invertible neural networks are especially useful. A conditional INN can invert a detector simulation in terms of high-level observables, specifically for ZW production at the LHC. It allows for a per-event statistical interpretation. Next, we allow for a variable number of QCD jets. We unfold detector effects and QCD radiation to a pre-defined hard process, again with a per-event probabilistic interpretation over parton-level phase space.