1998/11/19 by N. Kuropatkin, N.P. Kuropatkin, Kuropatkin, N. +2
Physics and Astronomy · #Computational Physics (physics.comp-ph) #Data Analysis #FOS: Physical sciences #Nuclear Physics and Applications #Particle Detector Development and Performance #Particle physics theoretical and experimental studies #Statistics and Probability (physics.data-an) #physics.comp-ph #physics.data-an
paper · pdf · doi:10.48550/arxiv.physics/9811039
6 pages, 3 figures, Presented at International Conference of Computing in High Energy Physics 98 (CHEP98)
arxiv created 1998/11/19 · openalex publication_date 1998/11/19 · arxiv updated 2016/09/07 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
A Feed Forward Error Back Propagation Artificial Neural Network(ANN) algorithm is developed for electron/positron identification in a wide momentum region (10 - 300 GeV/c). The method was proposed for the Transition Radiation Detector of the E781 experiment at Fermilab. The package consists of two parts: - the program for the ANN training; - the particle classification subroutine. Both parts are built using the object oriented technique and C++ language. The particle identification algorithm is wrapped in FORTRAN closers to be used in the E781 off-line program. The package performance was tested in comparison with the likelihood ratio method using Monte Carlo generated data. Our study has demonstrated the excellent ability of the ANN to learn even small details of the detector response function. The ANN solution gives the same performance and behavior as the likelihood method when using Monte Carlo data with known detector parameters. It demonstrates that if trained with experimental data the package can provide a very good solution to the classification problem of e+/e- tracks.