journal menu
![[Figure 1]](vh5156fig1mag.jpg)
|
|
Figure 1
A schematic description of the analysis pipeline. The pipeline consists of three main steps: (i) preprocessing, (ii) parameter prediction via the neural network and (iii) postprocessing. Step (i) includes geometric and other experiment-specific corrections. The data are also normalized, transformed into qz space, interpolated and standardized. In step (ii), the preprocessed data are fed into a trained fully connected neural network that yields an initial guess for the thin-film parameters. During step (iii), this initial guess is used as starting parameters for a fast Levenberg–Marquardt fit that finds the nearest LMS minimum. |
![[Figure 1]](vh5156fig1.jpg)
 | JOURNAL OF APPLIED CRYSTALLOGRAPHY |
ISSN: 1600-5767
access
