Implementation of a Machine Learning Regression Algorithm for Energy Reconstruction of Neutrino-induced Particle Showers using a Scintillating Fibres Tracker at the SND@LHC

SHiP and SND@LHC are two burgeoning experiments, as part of CERN, designed to study novel neutrino and BSM physics. Machine Learning (ML) algorithms need to be developed to extract hit information detected by SciFi planes and reconstruct energies of particle showers generated within the Sampling Calorimeter. This thesis implements real-life scenario into an existing ML algorithm for SND@LHC for particle shower reconstruction. Current algorithms do not tackle the ‘ghost-hits’ problem , however due to their presence in the real detector, these need to be taken into account before an algorithm can find immediate application in the detector. The ghost hit coordinates were appended to existing hit data to observe the increase in energy bias. Additionally, to eliminate the problem of ghost hits altogether, new side-view images of the SciFi planes were produced and fed into the algorithm to observe the change in energy bias and to optimize the network. The ghost hits failed to be implemented in the network, however, the optimized network successfully managed to minimize/eliminate energy bias to a value of 0.06%. The elimination/minimization of ghost hits will lead to greater adaptability and robustness of this ML algorithm to be adapted by the SND@LHC.