view article
Figure 3
(a) Box plots of the absolute errors for 242 measured reflectivity curves for each of the three predicted parameters. The upper and lower edges of the boxes represent the first and third quartiles, respectively, with the horizontal line inside the boxes denoting the median. The blue boxes represent the error compared with the pure neural network predictions. The pink boxes represent the error after applying a simple LMS minimization using the neural network predictions as starting parameters. The green boxes show the error for the case when a qz shift optimization has been performed before the LMS fit. (b) The same box plots of the median error but this time as a percentage of the ground truth. All results were obtained for a training noise level of n = 0.3.

Journal logoJOURNAL OF
APPLIED
CRYSTALLOGRAPHY
ISSN: 1600-5767
Volume 55| Part 2| April 2022| Pages 362-369
https://doi.org/10.1107/S1600576722002230
Follow J. Appl. Cryst.
Sign up for e-alerts
E-alerts
Follow J. Appl. Cryst. on Twitter
Twitter
Follow us on facebook
Facebook
Sign up for RSS feeds
RSS