view article
Figure 3
(a) Prediction errors in the simulation environment using both randomly generated samples (dashed lines) and patterns taken from an actual measurement of gold particles (solid lines) across different filter levels. With increased filter scaling, the adaptive policy (red) starts to outperform significantly both the raster scan (blue) and a simple baseline which uses the squared previous measurement as the exposure distribution (orange). At high filter levels, the adaptive policy eventually approaches the theoretical limit (green). Each point shows the average prediction error over 1000 episodes. (b) Measurement of gold particles on the XRF beamline at SSRL, which was used to evaluate the trained agent under more realistic conditions.

Journal logoJOURNAL OF
SYNCHROTRON
RADIATION
ISSN: 1600-5775
Volume 33| Part 1| January 2026| Pages 162-169
https://doi.org/10.1107/S1600577525009592
Follow J. Synchrotron Rad.
Sign up for e-alerts
E-alerts
Follow J. Synchrotron Rad. on Twitter
Twitter
Follow us on facebook
Facebook
Sign up for RSS feeds
RSS