Disentangling Boosted Higgs Boson Production Modes with Machine Learning

Yi-Lun Chung^1*, Shih-Chieh Hsu², Benjamin Nachman³

¹Department of Physics, National Tsing Hua University, Hsinchu 300, Taiwan
²Department of Physics, University of Washington, Seattle, Washington 98195, USA
³Physics Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA

* Presenter:Yi-Lun Chung, email:s107022801@m107.nthu.edu.tw

High transverse momentum(p_T) Higgs boson produced by gluon-gluon loop (ggF) is a sensitive probe of physics beyond the Standard Model at the Large Hadron Collider. However, high p_T Higgs boson from ggF is contaminated by substantial other production modes: vector boson fusion, production of a Higgs boson in association with a vector boson, and production of a Higgs boson with a top-quark pair. In this study, we introduce two-stream convolutional neuron network (2CNN) for disentangling high p_T Higgs decaying to bottom quark pair from each production mechanism. The first stream of 2CNN processes global collider event images and the second stream of 2CNN focuses on local leading non-Higgs jet images. This novel approach makes the fractional contribution of ggF to be improved from 0.4 to 0.6 in the highly boosted region (1000 GeV < p_T < 1250 GeV). Additionally, the approach in this study has the potential to improve the precision for other Higgs production modes in extreme regions of phase space. Moreover, we will present the dependence study between Monte Carlo generators and deep neural network (DNN), so called universality of DNN. This work is submitted to PRD (arXiv:2009.05930).

Keywords: Machine Learning, Higgs Productions, Boost