Particle Detection and classification in the High granularity calorimeter of Compact Muon Solenoid Detector

Aleesha, K T (2022) Particle Detection and classification in the High granularity calorimeter of Compact Muon Solenoid Detector. Masters thesis, Indian Institute of Science Education and Research.

Text (MS dissertation of Aleesha KT(17MS206)) 17MS072_Thesis_file.pdf - Submitted Version Restricted to Repository staff only Download (4MB)

Abstract

The primary purpose of this research is to investigate particle showers and particle identification in the Compact Muon Solenoid (CMS) experiment’s high-granularity calorimeter (HGCAL) using two machine learning methods, namely Artificial Neural Network and Graph Neural Network. We used data consisting of 8 types of particle trajectories( muon, positron, gamma, pion+, proton, kaon+, neutron, and kaon0L) generated by a standalone GEANT4 simulation of a slice of HGCAL similar to the one used in the October 2018 22a CMS beam test to train these supervised learning models. We have constructed an ANN Model named dr E Model and its variations that use layer features from the data to identify particles. For our GNN model dataset, we built a graph called hardware graph based on the detector’s geometry. We constructed Graph Convolutional Network(GCN) Models and Graph Attention Network(GAT) Models with four node and edge feature combinations for the event subgraphs. When it comes to classifying particles into four subclasses (Muon, Electromagnetic particles, Charged Hadrons, and Neutral Hadrons), the dr E Model and its modifications work admirably, but when it comes to classifying particles into eight classes, it falls short. A few GCN and GAT Model variants are presently close to matching the performance of the dr E Model, but more work is needed to improve it.

Actions (login required)

View Item