Generation of Synthetic Density Log Data Using Deep Learning Algorithm at the Golden Field in Alberta, Canada

Table 4

Training options for the DNN.

Parameter	Condition
Training algorithm	One of the algorithms from Table 1
Activation function	Log-sigmoid (for both the first and second hidden layers)
Maximum number of epochs	1,000
Maximum number of validation failures	10
Ratio of training and validation data	85%, 15% (random selection)