Machine learning approach in the prediction of differential cross sections and structure functions of pion electroproduction in the resonance region

This paper presents results of differential cross section prediction made by fully connected neural network. The exclusive reaction of electron scattering on a proton target with neutral pion production: $e^{-}p \rightarrow e^{-}p{\pi}^{0}$ is under our investigation. The model’s predictions are based on five kinematic variables. The quality of algorithm operation was evaluated by comparison between experimental data and the neural network’s predictions on distributions of unpolarized structure functions and the dependence of the differential cross sections on kinematic characteristics. A good agreement between the neural network’s predictions and experimental data allows interpolating the experimental data in five-dimensional space reliably.