Nuclear data generation by machine learning, 1; application to angular distributions for nucleon-nucleus scattering

Watanabe, Shoto*; Minato, Futoshi   ; Kimura, Masaaki*; Iwamoto, Nobuyuki  

In order to increase the efficiency of nuclear data evaluation, we have tested a combination of a nuclear reaction model and machine learning algorithm. We calculated nucleon-nucleus elastic scattering angular distributions by using the nuclear reaction model code, and optimized the potential parameters of an optical model to reproduce experimental data by means of the Bayesian optimization. We present optimization cases with the single parameter and two or more parameters, and show that our framework gives the angular distributions which are in good agreement with the observed ones.