Table 4.

Structure and parameters of attention

Network Structure Layer Kernel parameters Output dimension
Input (None, 2, Ns)

Convolution/BN/ReLU (64, 9) (None, 64, Ns)

Max pooling (4) (None, 64, Ns/4)

Residual block1 (64, 7) (None, 64, Ns/4)

Max pooling (2) (None, 64, Ns/8)

Residual block1 (64, 7) (None, 64, Ns/8)

Max pooling (2) (None, 64, Ns/16)

Attention (64) (None, 64, Ns/16)

Residual block2 (128, 7) (None, 128, Ns/16)

Max pooling (2) (None, 128, Ns/32)

Residual block2 (128, 7) (None, 128, Ns/32)

Max pooling (2) (None, 128, Ns/64)

Attention (128) (None, 128, Ns/16)

Residual block3 (256, 5) (None, 256, Ns/64)

Max pooling (2) (None, 256, Ns/128)

Residual block3 (256, 5) (None, 256, Ns/128)

Max pooling (2) (None, 256, Ns/256)

Attention (256) (None, 256, Ns/256)

Residual block4 (300, 5) (None, 300, Ns/256)

Adaptive average pooling (None, 300)

Fc (None, NUE)

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.