Context-enriched identification of particles with a convolutional network for neutrino events

Abstract

Particle detectors record the interactions of subatomic particles and their passage through matter. The identification of these particles is necessary for in-depth physics analysis. While particles can be identified by their individual behavior as they travel through matter, the full context of the interaction in which they are produced can aid the classification task substantially. We have developed the first convolutional neural network for particle identification which uses context information. This is also the first implementation of a four-tower siamese-type architecture both for separation of independent inputs and inclusion of context information. The network classifies clusters of energy deposits from the NOvA neutrino detectors as electrons, muons, photons, pions, and protons with an overall efficiency and purity of 83.3% and 83.5%, respectively. We show that providing the network with context information improves performance by comparing our results with a network trained without context information.

Toward A Precision Era Of Neutrino Oscillation Physics: Liquid Argon Scintillation Detector Development for DUNE and Neutrino Oscillation Studies with NOvA

“…Three different methods were explored in the context of conducting WS separation in NOvA within the context of the ν e appearance analysis working group. One method was explored originally by a colleague, via a method using the context-enriched CVN (Prong CVN or CVN Prong) [54] discussed previously with respect to energy estimation in Chapter 4. The reconstructed clusters of hits are evaluated in a CVN network that includes information from the event and relevant to this specific cluster, and the output is then a particle identification classifier as opposed to an event classifier.…”

“…A second CVN instance [54] is used in the energy estimate for ν e (ν e ) events and is also used in some of the wrong-sign measurements to be presented in Chapter 5. This CVN uses additional information from the reconstructed clusters to perform particle ID on individual clusters as opposed to event identification.…”

Constraints on Neutrino Oscillation Parameters from Neutrinos and Antineutrinos with Machine Learning

“…Another technique employing CNNs is that of classifying individual particles produced by the neutrino interaction [94]. Identifying particles is useful for particle energy estimation and identification of final states for measuring neutrino interaction cross sections.…”

Study of Neutrino Oscillations in the NOvA Experiment

“…This is of importance for understanding the topology of such particles in the detector, and the results of this are applicable in the training of neural networks for the recognition of these particles. Event reconstruction is also improved, and neural networks are used more widely for this [14,15]. The data analysis techniques and statistical methods are also modified.…”