An optimal choice of proper kinematical variables is one of the main steps in using neural networks (NN) in high energy physics. Our method of the variable selection is based on the analysis of a structure of Feynman diagrams (singularities and spin correlations) contributing to the signal and background processes. An application of this method to the Higgs boson search at the Tevatron leads to an improvement in the NN efficiency by a factor of 1.5-2 in comparison to previous NN studies.
@article{arxiv.hep-ph/0302088,
title = {Optimized Neural Networks to Search for Higgs Boson Production at the Tevatron},
author = {E. Boos and L. Dudko},
journal= {arXiv preprint arXiv:hep-ph/0302088},
year = {2009}
}
Comments
4 pages, 4 figures, partially presented in proceedings of ACAT'02 conference