journal menu
![[Figure 4]](vh5156fig4mag.jpg)
|
|
Figure 4
A comparison of the testing loss calculated from a simulated test set (100 000 curves, red) and an experimental test set (242 curves, blue) for different levels of uniform noise n that were applied to the training data. For each noise level a separate model was trained. With increasing noise level, the loss from the simulated data increases linearly, while the loss from the experimental data shows a clear minimum at noise levels of 0.3–0.35. The error bars represent the standard deviation from five independent training repetitions. |
![[Figure 4]](vh5156fig4.jpg)
 | JOURNAL OF APPLIED CRYSTALLOGRAPHY |
ISSN: 1600-5767
access
