Machine learning applications on event reconstruction and identification for ISS-CREAM

Thumbnail Image

Files

ICRC2021_061.pdf (1.42 MB)

Links to Files

https://pos.sissa.it/395/061/pdf

Permanent Link

http://hdl.handle.net/11603/22095

Collections

UMBC Center for Space Sciences and Technology (CSST) / Center for Research and Exploration in Space Sciences & Technology II (CRSST II)
UMBC Faculty Collection
UMBC Physics Department

Author/Creator

Author/Creator ORCID

Date

2021-07-12

Type of Work

conference papers and proceedings
Text

Department

Program

Citation of Original Publication

Yu, Monong et al.; Machine learning applications on event reconstruction and identification for ISS-CREAM; 37th International Cosmic Ray Conference (ICRC2021), 12 July, 2021; https://pos.sissa.it/395/061/pdf

Rights

This item is likely protected under Title 17 of the U.S. Copyright Law. Unless on a Creative Commons license, for uses protected by Copyright Law, contact the copyright holder or the author.
Public Domain Mark 1.0
This work was written as part of one of the author's official duties as an Employee of the United States Government and is therefore a work of the United States Government. In accordance with 17 U.S.C. 105, no copyright protection is available for such works under U.S. Law.

Subjects

Abstract

A supervised machine learning algorithm is applied to the visual representations of the energy deposits in two orthogonal views of the calorimeter of ISS-CREAM. Convolutional Neural Networks (CNNs) backed by Tensorflow are used to calibrate the sampled energy of the calorimeter and reconstruct the total primary energy of cosmic rays (CR), as well as for CR identification. The CNN regression models are trained on detailed Monte Carlo simulated events reproducing the behavior of the ISS-CREAM instrument suite, and the results indicate that a calorimeter energy reconstruction resolution of as good as 25% is achieved. The energy sampled in the calorimeter is determined with a resolution as good as 8%. The CNN classification model can reach a CR identification accuracy of up to 93%. The CR primary energy reconstruction results from machine learning methods are consistent with a simple scaling of the sampled energy. The increased accuracy of this CNN energy reconstruction comes from the additional information of the longitudinal and lateral energy deposit profiles. This machine learning approach is widely applicable to a range of particle physics and astrophysics problems.