Table 8.

Depression Dataset: results of the Wilcoxon signed-rank test comparing EasyMKLFS with respect to the others. Smaller p-values mean an higher difference between the models and, in our case, the Bonferoni corrected p-value is $0.05/8=6.25\cdot {10}^{-3}$.

	Algorithm	p-value w.r.t. EasyMKLFS
Baseline	SVM	$8.6\cdot {10}^{-5}$
FS	SVM RFE	$3.8\cdot {10}^{-4}$
	SVM t-test	$1.2\cdot {10}^{-4}$
RF	Gray	$4.3\cdot {10}^{-5}$
	Pustina	$7.8\cdot {10}^{-4}$
MKL	SimpleMKL	$1.8\cdot {10}^{-4}$
	EasyMKL	$4.6\cdot {10}^{-4}$
FW	EasyMKL	$9.6\cdot {10}^{-3}$