(Starting PhD seminar)
The goal of PANDA (anti-Proton ANnihilation at DArmstadt) experiment at FAIR (Facility for Anti-proton and Ion Research) is to study strong interactions at low energies. A high-intensity beam of antiproton will impinge on a proton target in the momentum range of 1.5 to 15 GeV/c. It will provide unique opportunities to study previously unexplored aspects of Quantum Chromodynamics at intermediate energies. PANDA will use a software trigger which requires track reconstruction and event building with unprecedented precision in real-time.
Tracking particle trajectories in such a dense environment is a challenging reconstruction task. So far conventional algorithms have been successfully used to build tracks, however, these methods are not yet able to reconstruct events in real-time. In the near future, such high-intensity environments will demand new tools for event reconstruction which are easily parallelizable, can handle non-linearities due to complex geometries, are scalable and consume less computing resources.
Deep neural networks have the ability to meet these requirements and have been effectively used in the computer vision for image and pattern recognition tasks. However, their use in track reconstruction is rather new and efforts need to be invested in the development of these algorithms for this purpose.
The present doctoral project is intended to explore the feasibility of deep learning algorithms for online event reconstruction. In this talk, I will describe the deep learning methods which will be explored and how they can be used in PANDA. I will describe the benefits and the challenges connected to these algorithms.