Sparsity Preserving Discriminant Projections with Applications to Face Recognition

Table 1

The best recognition rate and the corresponding standard deviation of the seven algorithms under the different size of the training set on Yale ( is the training sample size).

Methods
SPDP	0.7137 (0.031)	0.8333 (0.036)	0.9125 (0.027)	0.9458 (0.031)	0.9619 (0.025)	0.9786 (0.029)	0.9916 (0.033)
LPDP	0.5674 (0.057)	0.7175 (0.045)	0.7807 (0.037)	0.8186 (0.039)	0.8679 (0.035)	0.8816 (0.029)	0.9066 (0.036)
PCA	0.4389 (0.027)	0.4895 (0.035)	0.5514 (0.037)	0.5838 (0.048)	0.6241 (0.038)	0.6561 (0.043)	0.6727 (0.046)
LDA	0.5354 (0.061)	0.6486 (0.052)	0.7222 (0.036)	0.7792 (0.047)	0.8132 (0.037)	0.8375 (0.040)	0.8613 (0.044)
LPP	0.5783 (0.041)	0.6814 (0.044)	0.7469 (0.036)	0.8025 (0.035)	0.8139 (0.027)	0.8244 (0.014)	0.8392 (0.018)
NPE	0.5635 (0.025)	0.6811 (0.019)	0.7455 (0.027)	0.7593 (0.023)	0.8112 (0.017)	0.8284 (0.025)	0.8463 (0.023)
SPP	0.5202 (0.038)	0.6425 (0.027)	0.7098 (0.033)	0.7471 (0.033)	0.7653 (0.026)	0.7827 (0.032)	0.8037 (0.035)