Table 1 The list of the tuned hyperparameters
Hyperparameter | Prob/pair/mixed matrix | One-hot matrix |
|---|---|---|
Number of convolution layers | 2, 4, 6 | 1 |
Kernel size for convolution | 2, 3, 5 | [8, 16], [4, 8, 12, 16], [2, 4, 6, 8, 10, 12, 14, 16] |
| Â | Â | [2, 4, 6, 8, 10, 12, 14, 16] |
| Â | Â | [2, 4, 6, 8, 10, 12, 14, 16] |
Number of kernels (1st convolution layer) | 16, 32, 64 | 64, 128, 256, 512 |
Number of kernels (2nd convolution layer) | 32, 64, 128 | not applicable |
Pooling method | Max pooling, average pooling | Â |
Number of units (1st fully connected layer) | 64, 128,256 | 128, 256, 512 |
Number of units (2nd fully connected layer) | 32, 64, 128 | not applicable |
Learning algorithm | Adam, SGD | Â |
Dropout rate | 0.7, 0.5 | Â |