End-to-end physics reconstruction or what if we can't reconstruct new physics at the CERN LHC?

Authors

  • Michael Andrews Department of Physics, Carnegie Mellon University, USA

Abstract

Rule-based particle reconstruction algorithms underpin almost all major physics results of the CMS collaboration at the CERN LHC. Yet no new physics has been discovered since the Higgs boson in 2012 despite grounds for the contrary. Could the limitations of rule-based particle reconstruction algorithms perhaps be responsible for the lack of new physics results? In this talk, we consider models of exotic physics whose experimental signatures would go unnoticed in the CMS detector due to the limitations of current rule-based particle reconstruction algorithms. We then propose a novel end-to-end physics reconstruction technique and show how it is able to probe such exotic regimes that were once thought to be inaccessible. The technique leverages deep learning methods to train directly on "raw" detector data in order to bypass bottlenecks in current rule-based algorithms. We show how this new technique allows a number of never-before-seen measurements to be performed, and speculate on others yet it may enable.

About the Speaker

Michael Andrews, Department of Physics, Carnegie Mellon University, USA

Michael Andrews is a PhD candidate at Carnegie Mellon University. He works on the CMS experiment at the CERN LHC. His research focuses on using modern deep learning methods to search for rare and challenging signatures of physics beyond the standard model. He is the technical lead for a group of researchers focused on applying end-to-end deep learning techniques to physics searches at CMS, and to the real-time detection of anomalies in the CMS EM calorimeter. He has also served as the operations manager for the CMS EM calorimeter.

Downloads

Published

2021-09-05

How to Cite

[1]
M Andrews, End-to-end physics reconstruction or what if we can’t reconstruct new physics at the CERN LHC?, Proceedings of the Samahang Pisika ng Pilipinas 39, SPP-2021-INV-3B-02 (2021). URL: https://proceedings.spp-online.org/article/view/SPP-2021-INV-3B-02.