Skip to main content

Table 4 Model comparison of linear and non-linear ANN models

From: Application of neural networks with back-propagation to genome-enabled prediction of complex traits in Holstein-Friesian and German Fleckvieh cattle

 

Linear ANN

Non linear ANN

Best non-linear ANN

 

r

r

r

German Fleckvieh bulls

   

Milk yield DYD

0.68 (0.0007)

0.52 (0.0016)

0.68 (0.0008)

Protein yield DYD

0.68 (0.0006)

0.53 (0.0011)

0.67 (0.0005)

Fat yield DYD

0.66 (0.0005)

0.56 (0.0008)

0.65 (0.0005)

Holstein-Friesian bulls

   

Milk yield DYD

0.60 (0.0006)

0.53 (0.0011)

0.58 (0.0008)

Protein yield DYD

0.59 (0.0009)

0.50 (0.0013)

0.57 (0.0009)

Fat yield DYD

0.57 (0.0009)

0.51 (0.0010)

0.56 (0.0009)

Holstein-Friesian cows

   

Milk yield YD

0.47 (0.0031)

0.44 (0.0040)

0.47 (0.0027)

Protein yield YD

0.37 (0.0033)

0.35 (0.0039)

0.35 (0.0032)

Fat yield YD

0.46 (0.0037)

0.39 (0.0049)

0.47 (0.0028)

  1. Compared are linear and non-linear ANN with 1 neuron in hidden layer and G matrix as input to the network and best non-linear ANN. DYD = Daughter yield deviation, YD = Yield deviation, r = average Pearson correlation coefficient of the cross-validation runs, variance of cross-validation runs is shown in brackets.