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

Table 5

Classification accuracy of different graphs with varying densities “salt and pepper” noise.

“Salt and pepper”	0	0.02	0.04	0.06	0.08	0.1
LRKNN	0.901786	0.881786	0.842857	0.831786	0.799286	0.760714
LR	0.620357	0.559643	0.524286	0.505	0.493214	0.47
SR-	0.811786	0.803571	0.778571	0.755357	0.735714	0.710714
SR	0.750714	0.738214	0.680357	0.6425	0.617143	0.61285
LLE-	0.751429	0.732857	0.655357	0.559286	0.496071	0.455357
LLE	0.779286	0.737143	0.670714	0.569286	0.508571	0.472143
KNNK	0.612487	0.551039	0.504207	0.466663	0.442531	0.43228
KNNB	0.645828	0.586118	0.536622	0.498804	0.484881	0.468962