Skip to main content

The Center for Bright Beams, A National Science Foundation Science and Technology Center

Turn-key constrained parameter space exploration for particle accelerators using Bayesian active learning

Plots showing Bayesian optimization (BO) sampling patterns depending on radial basis function (RBF) kernel length scales and using α(x) = σ(x) as the acquisition function. Blue points are initial samples and orange points are sampled during Bayesian optimization. a Length scales for both [x1, x2] set to 1. b Length scales for variables [x1, x2] set to [0.25, 1].

Accelerator operators often need to characterize the beam response with respect to a set of input parameters. This process is made difficult since the beam response can have varying sensitivities to individual input parameters. Furthermore, practical constraints often make measuring beam parameters difficult or impossible. As a result, simple parametric scans used to perform this characterization task are inefficient to entirely impractical. Our work demonstrates a novel algorithm that autonomously characterizes the beam response in an efficient manner, with limited prior knowledge of measurement sensitivities, while navigating a highly constrained measurement space.

This work contributes towards the goals of Theme 3: Beam Dynamics and Control which aims to develop novel methods for improving control of high brightness accelerators. Using these advanced control methods, CBB enables improved transport of beams once they are generated to free electron laser and collider applications.

This advancement solves an extremely common experimental or computational problem faced by those in the accelerator, physical, chemical and biological scientific fields. As such, it is applicable towards experimental design in many different situations, especially novel or poorly understood contexts. Scientists in the physics community will be able to greatly improve the efficiency and success of their experiments using this work.


Roussel, R., Gonzalez-Aguilera, J.P., Kim, YK. et al. Turn-key constrained parameter
space exploration for particle accelerators using Bayesian active learning. Nat
 12, 5612 (2021).