Back to Article
Development of a Hemodialysis Data Collection and Clinical Information System and Establishment of an Intradialytic Blood Pressure/Pulse Rate Predictive Model
Journal of Artificial Intelligence and Big Data
| Vol 5, Issue 2
Table 2. Summary of best performances for BP/PRprediction taskusing different activation functions in ANN models (based on MAE).
| Activation function | Training loss | Validation loss | Test loss |
| SELU | 9.54 | 11.17 | 11.56 |
| SoftPlus | 12.61 | 12.51 | 12.59 |
| ELU | 12.90 | 12.62 | 12.82 |
| ReLU | 10.90 | 14.89 | 15.51 |
| Leaky ReLU | 10.39 | 14.38 | 14.21 |
| PReLU | 9.51 | 14.67 | 15.85 |