Table 4. Hyberparameters of CNN architectures.
| Parameter | Value |
|---|---|
| Batch size | Dataset size |
| Number of Convolutional Layers | VGG16 has 13, VGG19 has 16, AlexNet 5 convlutional layers and ResNet can have a varying number of layers. LSTM uses sequential data. |
| Learning rate θ | 0.001 |
| Regularization srength β | 0.001 |
| Input shape | 128 |
| epoch | 1,000 |
| Activation function | ReLU, softmax, and sigmoid |
| Momentum | 0.9 |
| Optimizer | SGD(lr=learning_rate, momentum=momentum) |
| Loss function | categorical_crossentropy and binary_crossentropy |