Spectral Nonlinearly Embedded Clustering Algorithm

Table 4

Performance comparison of different algorithms for the out-of-sample clustering on ten datasets.

Dataset	Nyström SC	KM	SEC	KSEC	ESEC
Iris	73.6 ± 1.8	75.6 ± 1.9	63.0 ± 1.9	77.1 ± 2.7	79.8 ± 2.1
Glass	49.7 ± 0.9	57.4 ± 1.2	58.6 ± 1.6	58.9 ± 2.4	63.1 ± 2.5
Wine	45.2 ± 0.9	54.6 ± 2.1	49.7 ± 1.9	61.2 ± 2.3	63.7 ± 1.9
WPBC	52.2 ± 2.8	64.7 ± 1.5	73.5 ± 1.5	74.3 ± 1.6	76.2 ± 1.4
SpectfHeart	51.4 ± 1.9	66.8 ± 2.9	53.5 ± 1.7	72.2 ± 2.3	70.8 ± 2.6
USPS	48.3 ± 4.5	56.0 ± 2.2	61.5 ± 2.4	65.1 ± 1.6	62.3 ± 2.8
Yale	20.7 ± 2.3	23.0 ± 1.6	35.9 ± 1.7	42.5 ± 1.6	45.8 ± 1.9
COIL20	63.9 ± 2.4	61.6 ± 3.3	68.9 ± 2.5	72.4 ± 2.7	70.2 ± 3.1
Isolet	34.8 ± 2.7	39.2 ± 1.2	45.3 ± 2.9	66.0 ± 1.8	68.4 ± 1.5
ORL	43.5 ± 2.1	47.6 ± 1.8	55.2 ± 1.5	60.7 ± 1.9	58.7 ± 1.7