Algorithm 2.

Repeated Cross Validation

1:	Split dataset into test and cross validation subsets $𝓣$ and $𝓓$, respectively.
2:	for each replicate $r\leftarrow 1,R$ do
3:	Shuffle subset $𝓓$.
4:	for each $\text{fold}\leftarrow 1$, $K$ do
5:	Split $𝓓$ into training and validation subsets, ${𝓓}_1$ and ${𝓓}_2$.
6:	Estimate transformation parameters from the training set ${𝓓}_1$.
7:	Apply the transformation to ${𝓓}_1$, ${𝓓}_2$ and $𝓣$ to standardize/normalize data.
8:	for each λ do
9:	Initialize model parameters $\mathit{\beta }$ or $\mathit{\alpha }$ by fitting a L2 SVM using (2) or (7).
10:	for each s do
11:	Fit a sparse classifier with $k=\lfloor p(1-s)\rfloor$ active features.
12:	Evaluate prediction accuracy on subsets ${𝓓}_1$, ${𝓓}_2$ and $𝓣$.
13:	end for
14:	end for
15:	end for
16:	Evaluate average scores $S_{\text{CV}}\left({𝓓}_1\right)$, $S_{\text{CV}}\left({𝓓}_2\right)$ and $S_{\text{CV}}(𝓣)$ for each subset.
17:	Rank models by their augmented scores ($S_{CV}\left({𝓓}_2\right)$, $s$, $\lambda$).
18:	Identify and record the optimal hyperparameters s, k, and λ.
19:	end for