Table 9 The number of parameters produce based on 1D-CNNs architecture to show the computational complexity
From: AFibNet: an implementation of atrial fibrillation detection with convolutional neural network
Layer name | Output shape | Parameters |
|---|---|---|
Convolution 1 | (None, 2698, 64) | 256 |
Convolution 2 | (None, 2696, 64) | 12,352 |
Maxpooling 1 | (None, 1348, 64) | 0 |
Convolution 3 | (None, 1346, 128) | 24,704 |
Convolution 4 | (None, 1344, 128) | 49,280 |
Maxpooling 2 | (None, 672, 128) | 0 |
Convolution 5 | (None, 670, 256) | 98,560 |
Convolution 6 | (None, 668, 256) | 196,864 |
Convolution 7 | (None, 666, 256) | 196,864 |
Maxpooling 3 | (None, 333, 256) | 0 |
Convolution 8 | (None, 331, 512) | 393,728 |
Convolution 9 | (None, 329, 512) | 786,944 |
Convolution 10 | (None, 327, 512) | 786,944 |
Maxpooling 4 | (None, 163, 512) | 0 |
Convolution 11 | (None, 161, 512) | 786,944 |
Convolution 12 | (None, 159, 512) | 786,944 |
Convolution 13 | (None, 157, 512) | 786,944 |
Maxpooling 5 | (None, 78, 512) | 0 |
Flatten | (None, 39936) | 0 |
Dense | (None, 1000) | 39,936,000 |
Dense | (None, 1000) | 1,001,000 |
Class | (None, 1) | 1001 |
Total of parameters | Â | 45, 846, 329 |