Table 1.
Hyperparameter settings of convolution-capsule networks (Conv-CapsNet).
| Hyperparameter | Setting |
|---|---|
| L2 normalization term | 0.001 |
| Activation | Relu |
| Batch size | 148 |
| Iteration epoch | 300 |
| Learning rate of network | 0.001 |
| Optimizer | Adam |
| Filter | 32 |
| Kernel_size | 3 |
| Number of nodes in the hidden feature layer | 128 |
| Number of nodes in the PrimaryCaps layer | 64 |
| Routing time | 2 |
| Dimension of each capsule | 8 |
| Length of PrimaryCaps | 2 |
| Length of DigitCaps | 2 |