A Neural Network Model of a Quasi-Periodic Elliptically Polarizing Undulator in Universal Mode

TL;DR

This paper presents a machine learning model that predicts the polarization and energy of light in an accelerator, trained on simulated data and calibrated with limited measurements, enabling efficient device setting optimization.

Contribution

It introduces a transfer learning approach for modeling a quasi-periodic elliptically polarizing undulator, combining simulated and limited measured data for accurate predictions.

Findings

Effective transfer learning from simulation to measurement data

Fast prediction of light polarization and energy

Potential for optimized insertion device settings

Abstract

Machine learning has recently been applied and deployed at several light source facilities in the domain of Accelerator Physics. We introduce an approach based on machine learning to produce a fast-executing model that predicts the polarization and energy of the radiated light produced at an insertion device. This paper demonstrates how a machine learning model can be trained on simulated data and later calibrated to a smaller, limited measured data set, a technique referred to as transfer learning. This result will enable users to efficiently determine the insertion device settings for achieving arbitrary beam characteristics.