Muon Identification Using Deep Neural Networks with the Muon Telescope Detector at STAR

Abstract

The installation of the muon telescope detector opened new possibilities for studying dimuon production at STAR. However, backgrounds from hadron punch-through and weak decays of pions and kaons make the identification of primary muons challenging. In this paper we present a study of shallow and deep neural networks trained as classifiers for the purpose of muon identification using information from the muon telescope detector at STAR. The performance of shallow neural networks is presented as a function of the number of neurons in their hidden layer. A hyperparameter optimization for determining the optimal deep neural network classifier architecture is presented. The optimized deep neural network is compared with shallow neural networks, boosted decision trees, likelihood ratios, and traditional cut-based PID techniques. The superiority of the deep neural network based muon identification technique is demonstrated and compared with traditional PID through the measurement of the $ϕ$ meson and the $ψ(2S)$ in p+p collisions at $\sqrt{s}$ = 200 GeV. The deep neural network based PID simultaneously provides higher signal efficiency, signal-to-background ratio, and significance of the $ϕ$ peak compared to traditional PID techniques. Finally, a deep neural network assisted technique for measuring the muon purity in data is presented and discussed.