Table 2.
Hyperparameters of all models used to test the new loss function presented in Section 3.1.
| Deep Learning Architecture | Hyperparameters |
|---|---|
| Bi-LSTM | Number of layers: 2 |
| Layer 1 units: 100 Layer 2 units: 50 |
|
| Activation function: Leaky ReLU | |
| DNN | Number of layers: 6 |
| Layer 1 units: 100 Layer 2 units: 500 Layer 3 units: 100 Layer 4 units: 250 Layer 5 units: 12 Layer 6 units: 6 |
|
| Activation function: ReLU | |
| CNN1D | Number of layers: 2 |
| Layer 1 units: 64 Layer 2 units: 64 |
|
| Activation function: ReLU | |
| Filter size: 3 x Features | |
| Bi-GRU | Number of layers: 2 |
| Layer 1 units: 100 Layer 2 units: 50 |
|
| Activation function: Leaky ReLU |