Speaker
Lewis Boulton
Description
Plasma accelerators often constitute a high-noise environment with multiple, non-linear dependencies that make the setup and operation of such devices a difficult task. To address these challenges, Machine Learning methods have gained popularity in the field of plasma acceleration. In this contribution, we summarise the application of such techniques to the beam-driven plasma acceleration experiment FLASHForward at DESY, Hamburg. Examples include the automated tuning of the plasma stage via Bayesian Optimisation and the development of non-destructive, neural-network-based predictions of the resulting accelerated trailing-bunch spectra.
Author
Lewis Boulton
Co-authors
Advait Laxmidas Kanekar
(DESY/UHH)
Andreas Maier
(DESY)
Angel Ferran Pousa
Brian Foster
(DESY)
Dr
Carl A. Lindstrøm
(University of Oslo)
Felipe Peña
(University of Oslo and Ludwig Maximilian University of Munich)
Gregor Loisch
(Deutches Elektronen-Synchrotron DESY)
Harry Jones
(DESY)
Jens Osterhoff
(Lawrence Berkeley National Laboratory)
Dr
Jonas Björklund Svensson
(Lund University)
Jonathan Wood
(DESY)
Judita Beinortaite
(FLASHForward, DESY, UCL)
Maryam Huck
(DESY)
Mathis Mewes
(DESY)
Matthew Wing
(UCL)
Maxence Thévenet
(DESY)
Pau Gonzalez Caminal
(DESY, Universität Hamburg)
Philipp Burghart
(DESY/UHH)
Richard D'Arcy
(University of Oxford)
Sarah Schroeder
(DESY)
Stephan Wesch
(Deutsches Elektronen-Synchrotron DESY)
Tianyun Long
(DESY)
