Fig. (8) Prediction error of training set (top) and test set (bottom) for model 1.