A Novel Graph Constructor for Semisupervised Discriminant Analysis: Combined Low-Rank and -Nearest Neighbor Graph

Table 3

Classification accuracy of different graphs with different label rates on four databases.

Graphs	LRKNN	LR	SR-	SR	LLE-	LLE	KNNK	KNNB
ORL (20%)	0.739063	0.564375	0.695	0.659688	0.675938	0.694063	0.438489	0.432261
ORL (30%)	0.865	0.753214	0.824643	0.8175	0.830357	0.84	0.696483	0.712444
ORL (40%)	0.9275	0.8575	0.9075	0.915417	0.916667	0.917083	0.889171	0.880085
ORL (50%)	0.9615	0.929	0.9535	0.959	0.953	0.944893	0.941861	0.9615
YaleB (20%)	0.872188	0.436875	0.704063	0.613438	0.645	0.684688	0.421005	0.435515
YaleB (30%)	0.901786	0.620357	0.811786	0.750714	0.751429	0.779286	0.612487	0.645828
YaleB (40%)	0.915833	0.742917	0.875417	0.840833	0.832083	0.829167	0.741223	0.789611
YaleB (50%)	0.9385	0.7805	0.931	0.892	0.8835	0.8925	0.824813	0.902639
PIE (20%)	0.623333	0.521667	0.622333	0.594667	0.594667	0.578667	0.376174	0.464566
PIE (30%)	0.752692	0.667308	0.736923	0.733462	0.749231	0.698846	0.559724	0.674269
PIE (40%)	0.820909	0.761364	0.839091	0.839091	0.852273	0.815455	0.72197	0.846924
PIE (50%)	0.885	0.833889	0.888333	0.881667	0.867222	0.908269	0.855231	0.875
USPS (20%)	0.685	0.684583	0.664167	0.685417	0.675417	0.609583	0.389417	0.41954
USPS (30%)	0.722857	0.69381	0.711429	0.67381	0.72	0.702857	0.510963	0.580313
USPS (40%)	0.793667	0.788333	0.787778	0.793333	0.790556	0.78	0.744394	0.765773
USPS (50%)	0.830667	0.844667	0.835333	0.828	0.827333	0.849333	0.86419	0.862344