Machine learning based prediction of electron and X-ray properties for non-invasive measurements

By virtue of the stability of the LUX beamline it is possible to generate unpreceded amounts of experimental laser-plasma acceleration data which can be used for data-driven modelling. In this work we show how, using an ensemble of neural networks, a model of the relationship between laser and electron properties can be extracted. The method generates a versatile model that is resistant to noise and can be used for optimization and non-invasive measurements. Further, it can give valuable insights into how jitter in laser parameters effects electron stability, such information can be used as guidelines for future laser development.