Table 2 Optimal values of model parameters, a: for predicting EC, and b: for predicting TDS
a | |
|---|---|
Model | Type of activation function hyperparameter |
LOST | Number of hidden layers:8, fixed-rate learning: 0.01, size of the batch: 32, and epochs of training: 1000 |
GPRE | Kernel function: Gaussian function, \(\sigma_{f}^{2}\):2 and l:1 |
CNNE | learning rate:0.01, kernel size:2 and pooling size:1 |
LOST-GPRE-CNNE: | LOST (Number of hidden layers:5, fixed-rate learning: 0.01, size of the batch: 32, and epochs of training: 1000) CNN (Learning rate:0.01, kernel size:2, and pooling size:1) GPR (kernel function: Gaussian, \(\sigma_{f}^{{}}\):2 and l:1 |
b | |
|---|---|
Model | Parameter values |
LOST | Number of hidden layers:8, fixed-rate learning: 0.01, size of the batch: 32, and epochs of training: 2000 |
GPRE | Kernel function: Gaussian function, \(\sigma_{f}^{2}\):2 and l:1 |
CNNE | learning rate:0.01, kernel size:2 and pooling size:1 |
LOST-GPRE-CNNE: | LOST (Number of hidden layers:8, fixed-rate learning: 0.01, size of the batch: 32, and epochs of training: 1000) CNN (Learning rate:0.01, kernel size:2, and pooling size:1) GPR (kernel function: Gaussian, \(\sigma_{f}^{{}}\):2 and l:1 |