Supervised Learning based optics corrections in circular accelerators

by Elena Fol (Cern)

Tuesday 20 Apr 2021, 16:00 → 17:00 Europe/Berlin

459 (30b)

Recently, a new approach for optics corrections based on supervised machine learning and linear regression has been developed. By providing simulations of magnetic errors and introduced deviations from the nominal optical functions as training data, regression models capable to predict quadrupolar errors of the entire lattice can be generated. This method has been extensively studied on historical LHC measurements and simulations of various optics configurations. In addition, machine learning based corrections are performed on the high luminosity upgrade of the LHC (HL-LHC) simulations, demonstrating the potential of this method for future applications in circular accelerators.  We also present the application of autoencoder neural networks to denoising of measurements data and reconstruction of missing data points. The results and future plans for these studies will be discussed following a brief introduction to relevant ML concepts

DESY_seminar_Fol.pdf