Table 4 HOA-Tuned and fixed hyperparameters.
Hyperparameter | Description | Tuned by HOA |
|---|---|---|
Learning rate | Controls the step size during training | Yes |
Batch size | Number of samples per training iteration | Yes |
Optimizer type | Adam, RMSProp, SGD | Yes |
Dropout rate | Prevents overfitting in FC layers | Yes |
GRU hidden units | Determines GRU capacity | Yes |
Number of convolution filters | In AlexNet layers | No |
Kernel size | In convolution layers | No |
