Graph Regularized Nonnegative Matrix Factorization with Sparse Coding

Table 1

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

Method
NMF	72.8% ± 2.0%	78.8% ± 1.5%	82.6% ± 2.0%	85.0% ± 3.5%	88.0% ± 3.5%
LNMF	61.4% ± 8.0%	65.4% ± 6.0%	67.5% ± 4.0%	71.3% ± 3.9%	74.2% ± 4.0%
SNMF	73.6% ± 1.0%	79.3% ± 1.0%	82.8% ± 2.0%	85.3% ± 2.0%	89.0% ± 1.5%
GNMF	73.9% ± 1.0%	80.9% ± 1.0%	83.8% ± 2.0%	86.2% ± 1.0%	89.5% ± 1.0%
FMD-NMF	72.5% ± 1.0%	82.1% ± 2.0%	87.0% ± 1.0%	88.5% ± 1.5%	92.0% ± 2.0%
S-GRNMF_SC	73.2% ± 2.0%	83.3% ± 3.0%	89.0% ± 2.0%	90.4% ± 1.0%	92.5% ± 2.0%