Applying Machine Learning Methods to Laser Acceleration of Protons: Lessons Learned from Synthetic Data*

Researchers in the field of ultra-intense laser science are beginning to embrace machine learning methods for control and optimization of secondary particles and radiation. In this study we consider three different machine learning methods and compare how well they can learn from a synthetic data set for proton acceleration that we generated using a modification of the Fuchs et al. 2005 model. This allows us to compare the machine learning models to each other and to the intrinsic noise level that was added to the data. We also provide results on the computational performance and memory consumption of the machine learning methods, which are important considerations for quasi-real time operation of these methods on real experiments.