Aachen 2019 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

T: Fachverband Teilchenphysik

T 4: Deep Learning I

T 4.3: Talk

Monday, March 25, 2019, 16:30–16:45, H06

Application of Deep Learning to Heavy Flavour Jet Identification with the CMS Experiment — Xavier Coubez^1,2, Luca Mastrolorenzo¹, •Spandan Mondal¹, Andrzej Novak¹, Andrey Pozdnyakov¹, and Alexander Schmidt¹ — ¹RWTH Aachen University, Germany — ²Brown University, USA

Many physics analyses within the CMS experiment rely on the efficient identification of heavy flavour jets. Over the past few years, several algorithms have been developed to exploit the distinctive features of jets arising from heavy flavour quarks to distinguish them from those arising from light quarks. The CMS collaboration has recently shown that Deep Neural Networks (DNNs) can be used to achieve significantly higher efficiencies while tagging heavy flavour jets, compared to traditional Machine Learning approaches. In addition to standard b-tagging and c-tagging algorithms, Deep Learning has been implemented to develop flavour tagging algorithms specialized for boosted topologies, to aid physics analyses that focus on boosted regimes and heavy exotic particles. This talk focuses on new advances in the application of Deep Learning in heavy flavour jet identification at CMS as well as the performance measurements of tagging algorithms on CMS data.