Research Article
An Intelligent Fault Diagnosis Based on Adversarial Generating Module and Semi-supervised Convolutional Neural Network
Table 5
Ablation study on hyper-parameters.
| Hyper-parameters | Values | Accuracy of MRFCD-SSCNN (with 200 labeled samples) |
| Learning rate | 0.1 | 81.2% (±1.33) | 0.01 | 85.6% (±1.15) | 0.001 | 93.4% (±0.89) |
| Pool size | 55 | 81.9% (±1.02) | 33 | 87.1% (±1.56) | 22 | 93.4% (±0.89) |
| Dropout rate | 0.2 | 78.2% (±2.11) | 0.5 | 93.4% (±0.89) | 0.7 | 84.7% (±1.48) |
|
|