A novel multi-purpose variational clustering architecture applied to neutron ID within the GlueX BCAL

Giroux, James
Journal Title
Journal ISSN
Volume Title
Faculty of Science, University of Regina

Particle Identification plays a crucial role within the GlueX Barrel Calorimeter. This paper details the implementation of a novel Machine Learning architecture, which combines cutting-edge conditional generative models with clustering algorithms, capable of extracting both e cient and high purity data samples, while only relying on information from only one type of sample. We demonstrate the validity of our approach and highlight its use as a neutron detection device, emphasizing its ability to limit assumptions on background samples. This architecture is flexible and can be extended to multiple categories. Remarkably it can be deployed for a wide range of problems, e.g., anomaly detection and data quality control.

A Thesis Submitted in Partial Fulfillment of the Requirements for the Degree of Bachelor of Science (Honours) in Physics, University of Regina. 40 p.
Electromagnetic calorimeter., BCAL., XGBoost., TensorFlow., Keras., HDBSCAN., DBScan., Scikit Learn., Boosted decision tree., Conditional Sigma Variational Autoencoder.