A Semisupervised Feature Selection with Support Vector Machine

Table 4

Comparisons of test errors and their standard errors (in parenthesis) on benchmark data sets.

Dataset	SENFS		Spectral		DrSVM
	𝑙=10	𝑙=100	𝑙=10	𝑙=100	𝑙=10	𝑙=100
Digit1	15.1 (0.04)%	4.28 (0.12)%	17.7 (0.09)%	8.7 (0.15)%	21.35 (0.05)%	5.85 (0.16)%
USPS	21.17 (0.03)%	11.1 (0.24)%	29.87 (0.15)%	12.6 (0.31)%	31.8 (0.12)%	17.3 (0.3)%
COIL2	36.1 (1.04)%	13.6 (0.98)%	37.7 (1.3)%	16.9 (0.91)%	47.2 (0.8)%	31.2 (1.1)%
BCI	43.9 (0.23)%	33.8 (0.24)%	49.87 (0.15)%	42.1 (0.31)%	47.8 (0.12)%	36.7 (0.8)%
g241c	32.8 (1.14)%	24.9 (1.12)%	42.8 (1.29)%	32.7 (1.5)%	41.75 (1.1)%	26.0 (1.6)%
g241n	37.6 (0.83)%	24.3 (0.24)%	45.7 (0.3)%	33.6 (0.51)%	41.8 (0.32)%	25.7 (0.67)%