Table 3 The DNN model accuracy is evaluated based on activation functions and learning rates.
Learning rate | ReLU | Tanh |
---|---|---|
0.008 | 84.05 | 81.34 |
0.009 | 84.02 | 81.24 |
0.01 | 84.07 | 81.44 |
0.02 | 83.08 | 81.28 |
0.03 | 83.38 | 81.18 |
0.04 | 83.09 | 81.12 |
0.05 | 81.19 | 80.14 |
0.06 | 82.70 | 80.70 |
0.07 | 81.76 | 79.80 |