view article

Figure 4
(a) A schematic diagram of the variational auto-encoder (VAE). It consists of two convolutional neural networks as encoder and decoder, and the model is fitted by comparing the similarity between the input and decoder-generated 2D slices with additional regularization by assuming the latent parameter follows an N(0, 1) distribution. (b) The distribution of the 10 000 bootstrapped 2D average patterns in CLPCA space (grey). Red dots are selected average patterns along the melting sequence used for the VAE analysis. (c) CLPCA-2 plotted against gain for the selected patterns from panel (b) along the melting sequence.

Volume 9| Part 2| March 2022| Pages 204-214
ISSN: 2052-2525