Table 3 Hyperparameter configuration for the SRA-MLPR mode.

From: Leveraging multi-source data and teleconnection indices for enhanced runoff prediction using coupled deep learning models

SRA-MLPR Hyperparameter

Parameter value

LHK

JP

GD

ET

Activation function

ReLU

ReLU

ReLU

ReLU

Learning rate

0.001

0.0001

0.001

0.0006

Hidden layer

6

5

7

7

Optimization algorithm

L-BFGS

L-BFGS

L-BFGS

L-BFGS

Iteration count(Default)

15,000

15,000

15,000

15,000

Convergence tolerance

1*10− 7

1*10− 7

1*10− 7

1*10− 7

Back to article page