Speaker
Konstantin Androsov
(INFN Pisa (Italy))
Description
The reconstruction and identification of tau lepton in semi-leptonic (hereinafter referred to as hadronic decays) are crucial for all analyses with tau leptons in the final state. To discriminate the hadronic decays of tau from all 3 main backgrounds (quark or gluon jets, electrons, and muons), maintaining a low rate of misidentification (below 1%) and at the same time with high efficiency on the signal, the information of multiple CMS sub-detectors must be combined. Application of deep machine learning techniques allows exploiting the available information in a very efficient way. Introduction of a new multi-class DNN-based discriminator provides considerable improvement of the tau identification performance at CMS.
Primary author
Konstantin Androsov
(INFN Pisa (Italy))
