Skip to main content
. Author manuscript; available in PMC: 2024 Jan 1.
Published in final edited form as: J Comput Graph Stat. 2022 Dec 13;32(3):1097–1108. doi: 10.1080/10618600.2022.2146697

Table 2.

Partial solution path in 5-fold cross validation on the synthetic example with various active variables k. Reported metrics are averages over 5 folds with standard errors in parentheses. The highlighted row corresponds to results for the true sparsity level s = 0.996, or equivalently k = 2. Here λ = 1 is fixed.

Accuracy (%)
k Iterations Loss Support Vectors Train Validation Test
500 7 0.300 623 98 77 72
(0) (0.004) (1) (0) (2) (1)
375 210 0.301 622 98 77 73
(4) (0.004) (2) (0) (2) (1)
250 421 0.313 622 98 77 74
(7) (0.004) (1) (0) (2) (1)
125 599 0.351 619 97 79 76
(9) (0.004) (2) (0) (1) (1)
73 696 0.388 622 97 82 79
(11) (0.004) (1) (0) (2) (1)
43 747 0.421 625 96 85 82
(4) (0.005) (1) (0) (2) (1)
25 813 0.450 627 95 86 85
(4) (0.004) (1) (0) (2) (1)
15 820 0.471 626 95 87 88
(6) (0.003) (1) (0) (2) (1)
9 866 0.488 627 95 91 91
(9) (0.002) (1) (0) (2) (1)
5 885 0.503 628 96 94 93
(3) (0.002) (1) (0) (2) (0)
3 891 0.513 628 97 96 95
(2) (0.002) (1) (0) (1) (0)
Accuracy (%)
2 896 0.519 627 99 100 100
(3) (0.002) (1) (0) (0) (0)
1 946 0.747 640 76 73 76
(3) (0.002) (0) (0) (0) (1)