Table 3.
Tuning parameters used for the transfer learning models.
Training Parameters | InceptionV3 | Resnet-50 | Resnet-50-TCN | |
---|---|---|---|---|
Optimizer | (sgdm) | (sgdm) | (sgdm) | Adam (sgdm) |
Initial Learn rate | 0.05 | 0.001 | 0.001 | 0.1 |
Learn Rate Schedule | ‘piecewise’ | ‘piecewise’ | ‘piecewise’ | ‘piecewise’ |
Learn Rate Drop Factor | 0.4 | 0.7 | 0.7 | 0.5 |
Learn Rate Drop Period | 4 | 10 | 10 | 8 |
Max Epochs | 200 | 150 | 150 | 4 |
Mini Batch Size | 16 | 64 | 64 | 8 |
Verbose | 1 | 1 | 1 | 1 |
Verbose Frequency (iterations) | 50 | 50 | 50 | 100 |
Gradient Threshold Method | ‘L2norm’ | ‘L2norm’ | ‘L2norm’ | ‘L2norm’ |
Gradient Threshold | Inf | Inf | Inf | Inf |
L2 Regularization | 1 × | 1 × | 1 × |
|