Graph Construction for Charged Particle Tracking using Point Cloud Networks

Graph neural networks (GNNs) have shown great potential in solving the complex problem of charged particle tracking, which is an essential task in high-energy physics. Charged particle tracking entails reconstructing the trajectory of charged particles produced after particle collisions, and the scalability of GNNs could enable this task to be performed with high efficiency. The first step in applying GNNs to particle detector data is to construct edges between hits, which then enables a GNN to operate on the graph and generate predicted trajectories. The current process for graph construction involves constructing all possible edges first and then removing edges that violate manually set geometric constraints. This project aims to improve the accuracy and efficiency of this graph construction step by creating a point cloud network (PCN) with PyTorch Geometric to perform this task. PCNs are a recently developed class of neural networks that operate directly on point cloud data, and they could be used to perform graph construction without manually selected geometric constraints and with fewer edge constructions in total. This project also involves optimizing hyperparameters for the PCN model architecture and the training process by using Ray tune and Optuna.