Table 2 Parameter descriptions as well as ranges for the LSTM model.
From: TSWS: An observation-based streamflow dataset of Tianshan Mountains watersheds (1901–2019)
Parametric | Descriptions | Value range | Type |
|---|---|---|---|
Steplength | Lag days for meteorological data on streamflow | [0,7] | Integer |
numLstmLayers | Number of LSTM learning layers | [1,3] | Integer |
Dropout_prob | Loss rate of lost layers | [0,0.5] | Double |
SolverName | Solution Methods | {sgdm, rmsprop, adam} | Categorical |
InitialLearnRate | Initial learning rate | [1e-3,1] | Log |
L2Regularization | L2 regularisation factor (weight recession) | [1e-10,1e-2] | Log |
GradientThreshold | Gradient thresholds | [1e-4,3] | Log |
