Table 1 Hyperparameter space.
From: Real-time reconstruction of high energy, ultrafast laser pulses using deep learning
Name | Range | Parameter type | Phase | Fluence | Baseline |
|---|---|---|---|---|---|
Batch normalization | (Yes, no) | Choice | Yes | Yes | No |
Dropout | (0, 0.25) | Continuous | 0.175 | 0 | 0 |
Learning rate | (0.00001, 0.01) | Continuous (log) | 0.008 | 0.003 | 0.001 |
Learning rate decay | (0.5, 1) | Continuous | 0.98 | 1 | 1 |
Number of layers | (3, 20) | Discrete | 4 | 8 | 5 |
Number of nodes | (128, 512) | Discrete | 505 | 360 | 256 |
Optimizer | (Adam, SGD, RMSProp) | Choice | Adam | Adam | SGD |
