Page 246 - Artificial Intelligence in the Age of Neural Networks and Brain Computing
P. 246
6. Future Directions of Research 237
first
the elements, (B)
in 12 3-way
DL-architecture: of masks autoencoder An the for
B. 19 and 19. to specialized
and features size is
11.12A vector layer
Fig. input output
in 228 feature the
shown the the
as between reduces figure,
vector the In
features operation POOL 1 ) compresses the 19 outputs to 10 latent variables that form the input to the final (C) fully connected MLP-NN with 7 hidden neurons, for binary tasks.
228 convolution (MAX
the layer (AD-MCI-HC)
of the pooling
Extraction performs max classification
[32]. a
in (CONV1) Then, 3-way
presented layer features. and
method convolutional 57 of MCI-HC)
the vector AD-HC,
11.13 of a (A), a
FIGURE Flowchart stage outputting (AD-MCI, classification.
