MLProject

This project, developed as part of the Machine Learning in Physics class (Fall 2024), explores anomaly detection in high-energy physics datasets using a Graph Neural Network (GNN). The study utilizes the R&D datasets from the LHC Olympics 2020, containing simulated particle events characterized by kinematic variables such as transverse momentum (pT), pseudorapidity (η), and azimuthal angle (ϕ). The GNN, implemented in Python, was designed based on concepts taught by Prof. H. B. Prosper and models particle interactions and spatial relationships through graph convolutions. Training, validation, and testing were performed on balanced datasets, achieving a classification accuracy of 93.5% for background events and 67.6% for signal events, with a threshold of 0.968. The primary script, CNNProject.ipynb, includes the training implementation and results. Detailed explanations of the methodology and findings are provided in the accompanying report.