Neural networks to identify particles using topological properties of calorimeters

The present work describes a neural particle classifier system based on topological mapping of the segmented information provided by a high-energy calorimeter, a detector that measures the energy of incoming particles. The achieved classification efficiencies are above 97.50% for the higher energy particle beams, even when experimental data exhibit unavoidable contamination due to the particle beam generation process, what could jeopardize the classifier performance. Some deterioration in the performance for the lower energy range is also discussed. The reduction on the dimensionality of the data input space caused by the topological mapping may be very helpful when online implementation of the classifier is required.