Figure 1

Choosing the optimal regularization parameters. Figure showing the estimated generalization ability of the first pair of extracted features from regularized dual CCA, for different combinations of regularization parameters τ _x (regularization parameter for the copy number variables) and τ _y (regularization parameter for the gene expression variables). The generalization ability is estimated as the correlation between the projection of a test set onto the first pair of features extracted from a training set, averaged over three cross-validation folds. The highest value, = 0.877 is attained for τ _x = 0.9 and τ _y = 0.3.