Graph Regularized Nonnegative Matrix Factorization with Sparse Coding

Table 2

Optimal average recognition rates ±   on YALE database with different number of training samples of each person (dimensions = 50).

Method
NMF	62.8% ± 5.7%	64.4% ± 3.5%	67.7% ± 4.0%	76.3% ± 1.0%	80.8% ± 2.0%
LNMF	61.5% ± 1.0%	68.9% ± 1.0%	70.7% ± 1.5%	79.0% ± 2.0%	83.1% ± 2.0%
SNMF	63.6% ± 3.5%	65.3% ± 2.5%	70.0% ± 2.0%	77.7% ± 2.0%	82.2% ± 2.0%
GNMF	63.8% ± 2.9%	66.7% ± 2.0%	72.0% ± 2.0%	80.0% ± 1.5%	83.1% ± 2.0%
FMD-NMF	65.3% ± 2.0%	68.5% ± 1.8%	71.7% ± 3.0%	79.8% ± 3.0%	82.2% ± 3.0%
S-GRNMF_SC	67.0% ± 1.5%	70.2% ± 3.0%	73.3% ± 2.0%	83.3% ± 1.5%	85.4% ± 2.0%