Speaker
Description
Linear particle accelerators require outstanding synchronization of the subsystems to provide high quality results for temporally resolved experiments. Considering the laser-based synchronization system at the European X-Ray Free-Electron Laser, the optimization process involves searching for the global optimum in a high-dimensional parameter space whereby safety constraints are fulfilled. For this purpose, Bayesian optimization is iteratively applied on subspaces of the high-dimensional optimization problem to maintain numerical tractability. Furthermore, modifications to an existing safety mechanism are developed which show significantly improved convergence rates. This opens the applicability to even higher-dimensional problems with high iteration cost.
The developed algorithm was applied on the main branch of the optical synchronization system.
