Inverting cosmic ray propagation by convolutional neural networks

Abstract We propose a machine learning method to investigate the propagation of cosmic rays based on precisely measured spectra primary and secondary ray nuclei Li, Be, B, C, O from AMS-02, ACE, Voyager-1. train two convolutional neural networks. One network learns how infer source parameters energy rays, other network, which is similar former, has flexibility learn data with added artificial fluctuations. Together simulated generated by GALPROP , we find that both networks can properly invert process reasonably well. This approach be much more efficient than traditional Markov chain Monte Carlo fitting for deriving if users choose update confidence intervals new experimental data. Both trained are available at ( https://github.com/alan200276/CR_ML ).