Description
Current software for Monte Carlo particle tracing allow to read Monte Carlo Particle Lists (MCPL) generated by other software, usually a neutronics simulator. Currently, it is possible to perform statistical estimates of the multivariate distribution of the phase-space variables of neutrons from a MCPL file. In this work, we propose an alternative that learns the multivariate probability distribution by means of current generative models developed by the machine learning community. We present ways of sampling these models both for Vitess and McStas, and discuss the advantages and disadvantages of the proposed method.
Authors
José Robledo
(Jülich Centre for Neutron Science 2, Forschungszentrum Jülich)
Klaus Lieutenant
(HZB)
Norberto Schmidt
(Jülich Centre for Neutron Science (JCNS-2))
Dr
Paul Zakalek
(JCNS-2, Forschungszentrum Jülich)
Dr
Stefan Kesselheim
(JSC,FZJ)
