Applying the machine learning methods to determine the linear optics parameters in the ThomX collector ring

The linear optics parameters are one of the most significant properties of the beam, which are controlled at the particle accelerators. Classical methods of analysis, such as component-independent analysis, employ turn-by-turn readings of the beam position monitors. As an alternative to the component-independent analysis, machine learning and neural networks are proposed for determining the beam parameters. This approach relies on the same input data as classical algorithms. This work shows training and usage of the neural network for analysis of the data from the collector ring of the ThomX accelerator facility.