Robust Semi-Supervised Manifold Learning Algorithm for Classification

Table 2

The classification rates of the 9 methods on CMU PIE data set with different noise densities.

METHOD	Reverse-noise			Gauss-noise			noiseless
	1/6	1/8	1/12	0.02	0.05	0.1
LLE + NFL	0.6088	0.6163	0.5925	0.6213	0.6888	0.6738	0.6263
SS-LLE	0.6613	0.6763	0.7350	0.7025	0.7838	0.8450	0.7363
RSSML-LLE	0.7763	0.7250	0.7388	0.7315	0.8300	0.8613	0.7613
RLLPE + NFL	0.6263	0.6675	0.6438	0.6613	0.7263	0.7050	0.6538
SS-RLLPE	0.6375	0.6788	0.7138	0.7125	0.7488	0.7463	0.7075
RSSML-RLLPE	0.6963	0.6875	0.7275	0.7513	0.7838	0.7875	0.7075
LTSA + NFL	0.4488	0.4325	0.2825	0.2838	0.3488	0.4288	0.6900
SS-LTSA	0.8300	0.8050	0.8425	0.8525	0.8475	0.8288	0.8863
RSSML-LTSA	0.8550	0.8213	0.8438	0.8763	0.8688	0.8375	0.9013