Table 1.

Nonlinear settings: randomly select combinations of hyper-parameters to search over

Method	Error (%)	TPR-1	TPR-2	FPR-1	FPR-2	F-1	F-2
Setting 1
$(p_1=500,p_2=500,n_1=200,n_2=150)$
iDeepViewLearn	1.89 (0.47)	100.00	100.00	0.00	0.00	100.00	100.00
iDeepViewLearn on stacked data	4.00 (0.47)	100.00	100.00	0.00	0.00	100.00	100.00
Sparse CCA + SVM	6.10 (0.73)	100.00	90.00	0.11	0.01	99.51	94.69
Deep CCA + TS + SVM	35.61 (2.22)	11.10	11.30	9.88	9.86	11.10	11.30
MOMA	44.96 (1.70)	22.00	29.90	8.67	7.89	22.00	29.90
MOMA + SVM	30.47 (6.05)	22.00	29.90	8.67	7.89	22.00	29.90
Random Forest on stacked data	1.94 (0.60)	70.10	98.00	3.32	0.22	70.10	98.00
SVM on stacked data	28.07 (0.65)	–	–	–	–	–	–
Setting 2
$(p_1=500,p_2=500,n_1=6000,n_2=4500)$
iDeepViewLearn	1.26 (0.11)	100.00	100.00	0.00	0.00	100.00	100.00
iDeepViewLearn on stacked data	1.38 (0.08)	100.00	100.00	0.00	0.00	100.00	100.00
Sparse CCA + SVM	4.25 (0.15)	100.00	90.00	0.00	0.00	100.00	94.74
Deep CCA + TS + SVM	0.66 (0.13)	30.40	21.60	7.73	8.71	30.40	21.60
MOMA	12.77 (8.63)	76.30	89.90	2.63	1.12	76.30	89.90
MOMA + SVM	0.63 (0.08)	76.30	89.90	2.63	1.12	76.30	89.90
Random Forest on stacked data	0.66 (0.05)	100.00	100.00	0.00	0.00	100.00	100.00
SVM on stacked data	2.31 (0.15)	–	–	–	–	–	–
Setting 3
$(p_1=2,000,p_2=2,000,n_1=200,n_2=150)$
iDeepViewLearn	2.56 (0.78)	99.98	99.88	0.00	0.00	99.98	99.88
iDeepViewLearn on stacked data	2.86 (0.73)	99.98	99.65	0.01	0.04	99.98	99.65
Sparse CCA + SVM	4.86 (0.88)	100.00	97.50	0.08	0.02	99.63	98.66
Deep CCA + TS + SVM	29.91 (1.27)	10.30	11.20	9.97	9.87	10.30	11.20
MOMA	46.14 (2.44)	16.40	13.68	9.29	9.59	16.40	13.68
MOMA + SVM	35.46 (5.91)	16.40	13.68	9.29	9.59	16.40	13.68
Random Forest on stacked data	5.40 (1.02)	58.67	89.88	4.59	1.13	58.67	89.88
SVM on stacked data	28.57 (0.53)	–	–	–	–	–	–

TPR-1; true positive rate for ${\mathbf{X}}^{(1)}$. Similar for TPR-2. FPR-1; false positive rate for ${\mathbf{X}}^{(1)}$. Similar for FPR-2; F-1 is the F measure for ${\mathbf{X}}^{(1)}$. Similar for F-2. The highest F-1/2 is in. (The mean error of two views is reported for MOMA; MOMA + SVM means selecting features using MOMA and training an SVM on the selected features)