Research Article
Interpatient ECG Heartbeat Classification with an Adversarial Convolutional Neural Network
Table 1
The configuration for the encoder network.
| | Layer | Output size | Kernel size | Padding | Strides | Dilation rate |
| | Inputs | (3 × 128 × 2) | — | — | — | — | | Conv layer1 | (1 × 126 × 16) | (3 × 3) | Valid | 1 | 1 | | Residual block1 layer1 | (1 × 126 × 16) | (1 × 3) | Same | 1 | 1 | | Residual block1 layer2 | (1 × 126 × 16) | (1 × 3) | Same | 1 | 3 | | Spatioattention module | (1 × 126 × 16) | — | — | — | — | | Residual block2 layer1 | (1 × 63 × 64) | (1 × 3) | Same | 2 | 1 | | Residual block2 layer2 | (1 × 63 × 64) | (1 × 3) | Same | 1 | 3 | | Spatioattention module | (1 × 63 × 64) | — | — | — | — | | Residual block3 layer1 | (1 × 32 × 64) | (1 × 3) | Same | 2 | 1 | | Residual block3 layer2 | (1 × 32 × 64) | (1 × 3) | Same | 1 | 3 | | Global average pooling | (64) | — | — | — | — |
|
|
