Table 3 Detailed structural information of the BiLSTM-CNN-Attention model.
Layers | Input | Output | Activation function |
|---|---|---|---|
Bidirectional_1 (Bidirectional layer_1) | 2 × 23 | 256 × 1 | ReLU |
Bidirectional_2 | 256 × 1 | 1024 × 1 | ReLU |
Reshape | 1024 × 1 | 32 × 32 × 1 | \ |
Conv2D_1 (Convolutional 2D layer_1) | 32 × 32 × 1 | 16 × 16 × 1 | ReLU |
Conv2D_2 | 16 × 16 × 1 | 9 × 9 × 1 | ReLU |
Pooling2D_1 | 9 × 9 × 1 | 2 × 2 × 1 | \ |
Flatten layer | 2 × 2 × 1 | 4 × 1 | \ |
Attention layer | 4 × 1 | 64 × 1 | ReLU |
Full connected layer | 64 × 1 | 5 × 1 | Softmax |
