Table 6.
Performance of Type 1 and Type 2 depthwise separable convolution models for prediction of BCR
| Model | Layer type | Depth | No. params | Batch size | Class label | ROC AUC |
|---|---|---|---|---|---|---|
| Model 3 | DSC1 | 5 | 3.5 × 106 | 8 | BCR | 0.521 ± 0.063 |
| Model 4 | DSC1 | 5 | 3.2 × 104 | 8 | BCR | 0.557 ± 0.068 |
| Model 5 | DSC2 | 6 | 1.3 × 106 | 8 | BCR | 0.588 ± 0.131 |
| Model 6 | POOL-DSC1 | 5 | 1.0 × 105 | 16 | BCR | 0.546 ± 0.065 |
Here, depth is number of layers with learnable parameters, excluding the output single node layer.