Application of a Neural Network Simulation to the Modeling of the COMPTEL background

Authors

M Varendorff, Max-Planck-Institut für extraterrestrische Physik
D J. Forrest, University of New Hampshire - Main Campus
Mark L. McConnell, University of New Hampshire - Main CampusFollow
James M. Ryan, University of New HampshireFollow

Abstract

COMPTEL onboard the Compton Gamma Ray Observatory (CGRO) measures gamma-rays in the background dominated energy range from 750 keV to 30 MeV. The event rate of the background varies strongly with orbital parameters, such as cut-off rigidity, position of the earth in the field ofview and location of the spacecraft above the earth. Some, but not all, parameters can be modelled with analytic functions. A Neural Network Simulation with two hidden layers of 24 neurons each, ten input neurons and one output neuron, was applied to the data. For the training of the Neural Network an error backpropagation algorithm was used. The resulting model was tested on an independent dataset, and the relativeimportance of the orbital parameters on the background model was derived.

Department

Physics

Publication Date

1996

Journal Title

Astronomy and Astrophysics Supplement Series

Publisher

European Southern Observatory (ESO)

Document Type

Article

Recommended Citation

M. Varendorff, D. Forrest, M. McConnell, and J. Ryan. APPLICATION OF A NEURAL NETWORK SIMULATION TO THE MODELING OF THE COMPTEL BACKGROUND. 1996, Astron. Astrophys. Suppl., 120, C699.