Diagnostic methods play a critical role in understanding the properties of fast particle beams, which is essential for experimentally validating theoretical studies in laser–plasma interactions. Two-screen magnetic spectrometers are commonly used to simultaneously measure both the energy and angular electron distributions. In a magnetic field, electrons are deflected according to their energy, resulting in light signals detected by a scintillator. However, the analysis of the obtained data often involves solving complex multi-parameter problems, which typically require heuristic approaches and manual intervention. In this work, we propose a method for reconstructing the electron distribution using a deep neural network. Unlike existing methods, the proposed approach enables the automatic and simultaneous reconstruction of both the energy and angular electron distribution. Since the labeled experimental data are unavailable, synthetic data generated through numerical simulations, combined with data augmentation techniques, are used for training the neural network. The neural network achieved a cosine similarity of 0.79 between experimental data and data obtained through numerical simulation based on the predicted distribution.
Rodimkov et al. (Sun,) studied this question.