Learning to Remove Pileup at the LHC with Jet Images

We present the Pileup Mitgation with Machine Learning (PUMML) algorithm for pileup removal at the Large Hadron Collider (LHC) based on the jet images framework using state-of-the-art machine learning techniques. We demonstrate that our algorithm outperforms existing methods on a wide range of jet observables up to pileup levels of 140 collisions per bunch crossing. We also investigate what aspects of the event our algorithms are utilizing by understanding the learned parameters of a simplified version of the model.

[1]  G. Salam,et al.  Energy correlation functions for jet substructure , 2013, 1305.0007.

[2]  Andrew Haas,et al.  ATLAS Simulation using Real Data: Embedding and Overlay , 2017 .

[3]  Luke de Oliveira,et al.  Jet-images — deep learning edition , 2015, Journal of High Energy Physics.

[4]  M. Cacciari,et al.  The anti-$k_t$ jet clustering algorithm , 2008, 0802.1189.

[5]  G. Kasieczka,et al.  Deep-learning top taggers or the end of QCD? , 2017, 1701.08784.

[6]  M. Cacciari,et al.  SoftKiller, a particle-level pileup removal method , 2014, The European physical journal. C, Particles and fields.

[7]  L. Dery,et al.  Weakly supervised classification in high energy physics , 2017, Journal of High Energy Physics.

[8]  J. Cogan,et al.  Jet-images: computer vision inspired techniques for jet tagging , 2014, 1407.5675.

[9]  M. Cacciari,et al.  FastJet user manual , 2011, 1111.6097.

[10]  P. Harris,et al.  Pileup per particle identification , 2014, 1407.6013.

[11]  Patrick T. Komiske,et al.  Deep learning in color: towards automated quark/gluon jet discrimination , 2016, Journal of High Energy Physics.

[12]  Peter Skands,et al.  An introduction to PYTHIA 8.2 , 2014, Comput. Phys. Commun..