view article
Figure 3
A brief illustration of the absolute embedding neural network (NN) model. (a) The average pattern from Dragonfly. (b) A polar representation of the pattern. (c) A stack of 1D Fourier transform magnitudes along the angular axis for each radial bin. The odd frequency components (due to inversion symmetry) and the higher frequencies for signals at smaller radii have been removed. These represent the feature vectors for the neural network. (d) An example of using absolute CLPCA on a selected cubic subset of frames (red dots). The grey dots represent the embedding of the pattern averages from the whole dataset. (e) Training labels from CLPCA.

IUCrJ
Volume 9| Part 2| March 2022| Pages 204-214
ISSN: 2052-2525
https://doi.org/10.1107/S2052252521012707