Figure 1.
Structure of ANN. For the analysis of 25 SNPs, 50 input layer units were provided. The number of hidden layer units was changed from the usual 6 to 10, to optimize the ANN for the highest possible prediction accuracy. The output layer had only 1 unit. Because the ANN model has connection weight parameters, which depend on the number of connection units, analysis of 25 SNPs with 6 hidden layer units requires 306 connection weight parameters.