Münster 2017 – wissenschaftliches Programm
HK 40.6: Vortrag
Donnerstag, 30. März 2017, 15:15–15:30, F 3
Machine learning for the analysis of low-mass dielectrons on Run II data with ALICE — •Alex Chauvin for the ALICE collaboration — Excellence Cluster, Garching, Germany
Dielectron pairs are an experimental tool to investigate the Quark Gluon Plasma (QGP), which is expected to be created during ultra-relativistic heavy-ion collision. The measured electron-positron pairs are created at different stages of the evolution of the hot and dense medium and do not interact strongly with the latter. Hence, dielectron pairs can carry information to describe the space-time evolution of the system, thereby allowing us to investigate the predicted restoration of chiral symmetry.
However, photon conversions contribute largely to the background of the dielectron signal we are after. Whereas dielectron pairs at very low mass (< 100MeV) are created, photon conversion rejection leads to systematic uncertainty in the crucial mass range used for normalisation and extraction of virtual photons. The Toolkit for Multi-Variable Analysis allows us to consider several variables with different classification methods, such as Boosted Decision Trees, while obtaining a higher signal efficiency.
In this talk we will present the advantages of using machine learning for background rejection and how this method preserves signal efficiency. To illustrate it, we will further apply the method on the Run II data recorded by the ALICE experiment.
This work is supported by BMBF-FSP 202 and the Excellence Cluster Universe.