Vertex finding in neutrino-nucleus interaction: a model architecture comparison

Akbar F.; Ghosh A.; Young S.; Akhter S.; Ahmad Dar Z.; Ansari V.; Ascencio M.V.; Sajjad Athar M.; Bodek A.; Bonilla J.L.; Bravar A.; Budd H.; Caceres G.; Cai T.; Carneiro M.F.; Díaz G.A.; Felix J.; Fields L.; Filkins A.; Fine R.; Gaur P.K.; Gran R.; Harris D.A.; Jena D.; Jena S.; Kleykamp J.; Klustová A.; Last D.; Lozano A.; Lu X.-G.; Maher E.; Manly S.; Mann W.A.; McFarland K.S.; Messerly B.; Miller J.; Moreno O.; Morfín J.G.; Nelson J.K.; Nguyen C.; Olivier A.; Paolone V.; Perdue G.N.; Plows K.-J.; Ramírez M.A.; Ruterbories D.; Su H.; Syrotenko V.S.; Waldron A.V.; Yaeggy B.; Zazueta L. · Journal of Instrumentation (2022) · DOI: 10.1088/1748-0221/17/08/t08013

Resumen

We compare different neural network architectures for machine learning algorithms designed to identify the neutrino interaction vertex position in the MINERvA detector. The architectures developed and optimized by hand are compared with the architectures developed in an automated way using the package “Multi-node Evolutionary Neural Networks for Deep Learning” (MENNDL), developed at Oak Ridge National Laboratory. While the domain-expert hand-tuned network was the best performer, the differences were negligible and the auto-generated networks performed as well. There is always a trade-off between human, and computer resources for network optimization and this work suggests that automated optimization, assuming resources are available, provides a compelling way to save significant expert time.

Instituciones: Pontificia Universidad Católica del Perú