Page 258 - PVT Property Correlations

Artificial Neural Network Models for PVT Properties Chapter | 10 227

I1 Weights Transfer Activation
W function function
Input nodes I2 I × W I × W O activation
Output
after
I3
After
transformation
I4 before
activation
Bias B Threshold C
node
FIGURE 10.2 Transformation and activation functions in an ANN. ANN, artificial neural
network.

FIGURE 10.3 Single-layer ANN.
ANN, artificial neural network.

treated with an activation function to compute the output value. An option in
the transformation calculation is to use a bias node for the first hidden layer
nodes. The bias node value or bias factor scales up the weighted sum to its
value. This bias factor guarantees that the values in the input nodes are main-
tained for subsequent calculations. Another optional control is the activation
threshold. The activation threshold is a value used to guarantee that the cal-
culated output from the node is higher than or equal to the threshold value,
ensuring that only significant outputs are carried to the next layer.