Skip to main content

Table 1 Hyperparameters and corresponding values that are tested during the random search

From: An operational strategy for district heating networks: application of data-driven heat load forecasts

Hyperparameter Tested values
Scaling { None, Z-Score, Min-Max Scaling }
Training algorithm { SGD, AdaGrad, RMSProp, Adam }
Activation function { Sigmoid, ReLU, tanh, linear (in the output layer) }
Hours of input data { 24 × 3, 24 × 5, 24 × 7, 24 × 9 }
Learning rate {lrd×10−1,lrd,lrd×101,lrd×102} with (lrd) = default learning rate of the
  corresponding optimiser as implemented in the python keras api
Hidden layers {1,2,3,4}
Decay {0,0.0001,0.001,0.01}
Patience of early stopping {10,20,30}
Test split {0.25,0.3,0.35}
L2Regularisation λ{ 0, 0.001, 0.01, 0.1}
Dropout { 0.1, 0.2, 0.3 }