Figure 4.

(a) The internal construction of an encoder. The encoder is connected by N coding layers with the same structure, where N is set to 4. Each encode layer is composed of two sub-layers. The first sub-layer is a multi-head attention mechanism [25] and here it has four heads. The second sub-layer is a feed-forward neural network. A residual connection [26] is used to connect the two sub-layers, followed by a layer normalization [48]. (b)The internal structure of the densely connected convolutional neural network block is the so-called DC-CNN block. Conv1D means one-dimensional convolution. The output sequence of the encoder is converted into a group of sequence feature maps by the densely connected convolution operation. Intra-BCLs between two convolutional layers in each DC-CNN block are used to connect the previous and current feature maps [27].