Influence of QCD parton shower in deep learning invisible Higgs through vector boson fusion

TL;DR

This paper investigates how QCD parton shower modeling affects deep learning analyses of invisible Higgs decays via vector boson fusion at the LHC, emphasizing the importance of accurate theoretical projections for neural network performance.

Contribution

It provides a detailed study of the impact of QCD parton shower schemes on deep learning-based Higgs invisible decay searches, linking theoretical modeling with neural network effectiveness.

Findings

Different recoil schemes influence neural network sensitivity.

Higher-order QCD computations improve radiation pattern modeling.

Accurate parton shower modeling enhances deep learning analysis.

Abstract

Vector boson fusion established itself as a highly reliable channel to probe the Higgs boson and an avenue to uncover new physics at the Large Hadron Collider. This channel provides the most stringent bound on Higgs' invisible decay branching ratio, where the current upper limits are significantly higher than the one expected in the Standard Model. It is remarkable that merely low-level calorimeter data from this characteristically simple process can improve this limit substantially by employing sophisticated deep-learning techniques. The construction of such neural networks seems to comprehend the event kinematics and radiation pattern exceptionally well. However, the full potential of this outstanding capability also warrants a precise theoretical projection of QCD parton showering and corresponding radiation pattern. This work demonstrates the relation using different recoil schemes in the parton shower with leading order and higher-order computation.