Nikita Koloskov

Mentors: Giuseppe B. Cerati & Jeremy Hewes & Daniel Grzenda

Project Title: Exa.TrkX: Improving Graph Neural Network Performance for Classifying Neutrino Interactions in MicroBooNE Data

Project Description: The Exa.TrkX project presents a graph neural network (GNN) technique for low-level reconstruction of neutrino interactions in a Liquid Argon Time Projection Chamber (LArTPC). We discovered that the GNN model trained on DUNE simulation data performs quite poorly on the data from another neutrino detection experiment (MicroBooNE). After my colleague Kaushal modified the model architecture that allowed to detach the physical meaning from neutrino interaction graph edges, I explored new edge-forming techniques (such as Delaunay triangulation, KNN-graph, and radius graph) and retrained the model on MicroBooNE data, which resulted in 80% classification accuracy for physically meaningful interactions.