Demonstration of machine learning-enhanced multi-objective optimization of ultrahigh-brightness lattices for 4th-generation synchrotron light sources

A machine-learning based closed orbit feedback for a synchrotron light source

“…Nowadays, machine learning is more and more widely used in the accelerator field. For examples, researchers at SLAC replaced the traditional genetic and particle swarm algorithms with a machine learning-based method to optimize the nonlinear problem of the storage ring and find the multi-objective optimized Pareto front, and there is also a recent publication using machine learning to improve the search for the Pareto-optimal front for a fourth-generation storage ring [18,19]. Researchers at LBNL extended the traditional feed forwards method with machine learning algorithms to counteract the insertion device gap or phase motion-induced perturbations on the ALS light source electron beam [6].…”

“…Y. Lu et al have proposed the idea of using ML-MOGA as a full-fledged replacement for the traditional ML-enhanced MOGA in the optimization of 4GSR lattices. Their research has opened up new possibilities for huge speedups of traditionally lengthy computational processes [19,21].…”

Linear Optics Calibration in a Storage Ring Based on Machine Learning

“…Currently, machine learning is increasingly being used at accelerators worldwide [14,15]. For example, machine-learning-based methods are used to optimize nonlinear storage ring problems and to find the multiobjective optimized Pareto front [16,17]. A study at Lawrence Berkeley National Lab (LBNL) has shown that machine learning algorithms can be used instead of traditional feedforward methods to counteract the insertion device gap or phase motion-induced perturbations on the advanced light source (ALS) light source electron beam [6].…”

“…In addition, researchers from the Institute of High Energy Physics have proposed a new dynamical aperture (DA) prediction method based on machine learning, which can reduce the computational cost of DA tracking by approximately an order of magnitude while maintaining sufficiently high evaluation accuracy [19]. Y. Lu et al proposed using ML-MOGA as a full-fledged replacement for the traditional Tr-MOGA in optimizing 4GSR lattices [17]. Their research opened up new possibilities for considerable speedups of traditionally lengthy computational processes.…”