Identifying Isolated Photons in relativistic heavy ion collisions using Convolutional Neural Networks

Isolation cuts are crucial for accurate particle analyses in high-energy physics, including the measurement of photons from Compton processes and Z-boson decays. Identifying isolated particles becomes particularly challenging in heavy ion collisions due to the complex and dense background events. Ideally, an "isolated particle" is one that does not belong to any jet. This study focuses on designing, developing, and implementing a highly accurate Convolutional Neural Network (CNN) to distinguish isolated photons from the overwhelming background of neutral pion decays within jets. The CNN was trained on a dataset with photons categorized into two groups: those from jets and truly isolated photons. The successful application of this model enhances the identification of isolated photons in collisions at the Large Hadron Collider beauty (LHCb) experiment at CERN and provides a framework for developing CNN models to identify other isolated particles in high-energy collisions, supporting the exploration of new and unexpected particle physics phenomena.