Study Cosmic Ray Mass Composition Using Deep Learning in Telescope Array Surface Array Detector

The ultra-high-energy cosmic rays mass composition study with the Telescope Array surface detector is discussed. We present the new analysis based on deep convolutional neural network using detector signal time series as an input and trained on a large Monte-Carlo dataset. We compare the sensitivity of the new technique and the previously presented boosted decision tree multivariate analysis built upon 14 observables. Possible systematic errors of the method are discussed.