| 1 |
Training loss function |
Categorical Crossentropy, Categorical Hinge, KLDivergence, Poisson, Squared Hinge, and Hinge |
| 2 |
Training batch size |
4 → 48 (step = 4) |
| 3 |
Model dropout ratio |
[0 → 0.6] |
| 4 |
Transfer learning freezing ratio |
1 → 100 (step = 1) |
| 5 |
Weights (i.e., parameters) optimizer |
Adam, NAdam, AdaGrad, AdaDelta, AdaMax, RMSProp, SGD, Ftrl, SGD Nesterov, RMSProp Centered, and Adam AMSGrad |
| 6 |
Dimension scaling technique |
Normalize, Standard, Min Max, and Max Abs |
| 7 |
Utilize data augmentation techniques or not |
[Yes, No] |
| 8 |
The value of rotation (In the case of data augmentation). |
0° → 45° (step = 1°) |
| 9 |
In the case of data augmentation, width shift value. |
[0 → 0.25] |
| 10 |
The value of height shift if data augmentation is applied |
[0 → 0.25] |
| 11 |
Value of shear in case of data augmentation |
[0 → 0.25] |
| 12 |
Value of Zoom (if data augmentation is used) |
[0 → 0.25] |
| 13 |
Flag for horizontal flipping (if data augmentation is utilized) |
[Yes, No] |
| 14 |
(If augmentation of data has been applied), the value of Vertical flipping flag |
[Yes, No] |
| 15 |
Range of brightness changes (if data augmentation is applied) |
[0.5 → 2.0] |
