Aachen 2019 – scientific programme

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

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 2: Machine-learning methods and computing in particle physics

AKPIK 2.4: Talk

Tuesday, March 26, 2019, 16:30–16:40, H10

Learning to rank Higgs-Boson candidates — Lukas Pensel, Alexander Segner, •Marius Köppel, Martin Wagener, Andreas Karwath, Christian Schmitt, and Stefan Kramer — Johannes Gutenberg University Mainz, Germany

The Higgs boson was discovered in July 2012 and the Nobel prize was awarded for its theoretical prediction of this boson subsequently in 2013. The current research focus is now on the precise measurement of the Higgs boson couplings and the search for possible deviations from the predictions by the standard model of particle physics in as many production and decay channels as possible. One of the interesting channels is the production of the Higgs Boson via vector boson fusion, with subsequent decay H → WW^∗ → l ν l ν. Since this channel, as well as many others, is limited by large background contributions, in this case dileptonic t t decays, maximizing the signal to background ratio is a crucial step in these analyses. Commonly, machine learning models using boosted decision trees and standard neural networks are trained to differentiate these signals from background. Machine learning approaches can also be employed for ranking problems aiming to sort a list according to the relevance labels of its elements. This can be done by pairwise comparison of all instances in the list. Recently, neuronal networks have been used for this task as well. Here, we present a method using such a pairwise approach from the field of learning to rank combined with neuronal networks to sort a list of Higgs-events, depending on their relevance class, into signal and background.