Description
Searches for the new phenomena that consider signature with $\tau$ leptons is of big interest at the CERN LHC. The reconstruction of tau decaying to hadrons ($\tau_h$) is based on the hadron-plus-strip (HPS) algorithm, which combines the charged hadrons and $\pi^0$ candidates, obtained by clustering photon and electron candidates. However, in contrast to this combinatorial approach this problem can be solved with using constituent particles (electrons, photons, charge hadrons) and its properties as an input to the neural network to regress the visible component of $\tau_h$ momentum. This project is aimed at studying of applicability of different neural network architectures and input representations for the tau lepton reconstruction. Such flexible algorithm would give an opportunity to optimize tau reconstruction in the regions where current HPS algorithm has poor performance, for instance significantly displaced taus.
Special Qualifications:
Basic knowledge of Particle Physics, good knowledge of python, experience with neural networks and Linux system.
Field | B1: Particle physics analysis (software-oriented) |
---|---|
DESY Place | Hamburg |
DESY Division | FH |
DESY Group | CMS |