Exploring Deep Learning Methods for Particle Track Reconstruction

The fundamentals of particle are studied at CERN by smashing beams of protons inside a particle accelerator. The particle coming out of the collision fly all around the space leaving charge on the detector. The trajectory of particles coming out of collision needs to be reconstructed(from data gathered from detectors) in order to identify the particle type, and to study the properties of the particle, this process is called Particle Track Reconstruction (PTR). PTR is a challenging task in high energy physics as the rate of data generated at CERN has already reached tens of petabyte per annum. Further, with the advancement of Large Hadron Collider (LHC) to High-Luminosity Large Hadron Collider (HL-LHC) by the next decade, the data generated is expected to increase around 10 times due to increase in the rate of collision. So, there is a need for a fast, accurate and scalable algorithm for PTR. Deep Learning is a popular machine learning technique that has already shown promising results in the area of computer vision, natural language processing and modelling non-linear dependencies. In this paper, the authors present four different deep learning methods for reconstructing particle track and discuss the strength and weakness of each method by evaluating them on a simulated dataset of collision between protons in the LHC.