Table 3 Baseline results for cancer detection, using a vanilla AE with random weights
From: Using autoencoders as a weight initialization method on deep neural networks for disease detection
Top Layers (AE) | Accuracy (%) | MCC | Precision (%) | Recall (%) | Fscore (%) | |
|---|---|---|---|---|---|---|
Thyroid | AE: Encoding Layers | 83.03 ±2.17 | 0.16 ±0.16 | 52.46 ±38.70 | 15.10 ±18.39 | 18.80 ±19.71 |
AE: Complete AE | 93.07 ±1.52 | 0.76 ±0.04 | 81.12 ±9.41 | 79.58 ±8.46 | 79.57 ±3.87 | |
Skin | AE: Encoding Layers | 82.87 ±2.77 | 0.23 ±0.10 | 43.46 ±10.55 | 25.00 ±9.75 | 30.98 ±9.73 |
AE: Complete AE | 87.47 ±4.28 | 0.54 ±0.06 | 64.80 ±12.79 | 59.79 ±9.38 | 60.55 ±5.11 | |
Stomach | AE: Encoding Layers | 84.63 ±2.41 | 0.19 ±0.06 | 42.11 ±9.80 | 17.33 ±7.71 | 22.90 ±7.37 |
AE: Complete AE | 87.40 ±2.68 | 0.47 ±0.10 | 55.77 ±10.29 | 51.66 ±8.72 | 53.24 ±8.33 | |
Breast | AE: Encoding Layers | 82.13 ±4.16 | 0.22 ±0.10 | 53.51 ±20.92 | 20.60 ±12.96 | 25.94 ±13.06 |
AE: Complete AE | 87.00 ±1.58 | 0.52 ±0.04 | 62.81 ±7.03 | 57.80 ±6.29 | 59.70 ±3.35 | |
Lung | AE: Encoding Layers | 81.60 ±1.26 | 0.15 ±0.07 | 40.78 ±11.11 | 14.88 ±8.74 | 20.45 ±9.95 |
AE: Complete AE | 85.30 ±3.50 | 0.50 ±0.06 | 59.78 ±11.76 | 59.11 ±9.72 | 57.99 ±4.88 |