Table 1.
Optimal-performing CNN and FCN Architectures layout along with their corresponding hyperparameter configurations.
| Layers | CNN | FCN |
|---|---|---|
| Convolutional | Neurons: 1024 | Not Applicable |
| Kernel Size: 5 | ||
| Activation Function: ReLU | ||
| Pooling | Padding: 0 | Not Applicable |
| Strides: 1 | ||
| Flattening | None | Not Applicable |
| Fully Connected | Neurons:512 | Neurons: 512 |
| Activation Function : ReLU | Activation Function: ReLU | |
| Dropout: 0.2 | Dropout:0.2 | |
| Kernel size: 3 | Kernel size: 3 | |
| Output | Activation Function: tanh | Activation Function : ReLU |
| Learning rate: 6e-7 | Learning rate: 6e-7 | |