Many, if not all, ATLAS analyses rely on particle identification and reconstruction. However, that is
not an easy task and misidentification is not only possible but a regular occurrence. For that reason,
improvements of identification algorithms must be made. One such improvement targets photon
identification. The default photon identification at ATLAS uses simple
rectangular cuts on specific variables to identify photons and separate them from background sources.
Various early tests with machine learning techniques show promising results,
but further work on these techniques is required.
The thesis' topic would include the optimisation and validation of different neural network models
in order to improve the photon identification,
while simultaneously lowering the associated uncertainties.
A basic knowledge of Python and artificial neural networks is helpful but not required.
Wird geladen