Table 3.
CNN-MNIST: the topology of the CNN for handwritten digit classification.
Layers | Parameters | Dimensions |
---|---|---|
Input | — | (1,28,28) |
Convolution | (8,3,3) | (8,28,28) |
Activation (sigmoid) | — | — |
Average pooling | (2,2) | (8,14,14) |
Convolution | (16,3,3) | (16,14,14) |
Activation (sigmoid) | — | — |
Average pooling | (2,2) | (16,7,7) |
Flatten | — | (784) |
Fully connected | 100 | (100) |
Activation (sigmoid) | — | — |
Fully connected | 10 | (10) |
Activation (softmax) | — | — |