Exact -Component Graph Learning for Image Clustering

Table 2

Clustering performance.

Methods	-means	R-cut	ST-SC	RG-SC	CAN	CLR	-GC
ACC
Two-moon	75.50  0.00	73.00  0.00	79.00  0.00	78.00  0.00	93.00  0.00	87.50  0.00	98.50  0.00
Path-based	74.33  0.00	76.85  2.05	79.33  0.00	60.66  0.00	87.00  0.00	87.00  0.00	99.00  0.00
Yale	50.30  2.68	29.75  1.86	57.63  1.49	62.90  0.25	52.73  0.00	59.39  0.00	68.48  0.00
ORL	55.45  3.10	30.12  1.37	68.27  1.78	59.82  0.86	56.50  0.00	63.75  0.00	70.75  0.00
COIL-20	66.79  2.65	45.72  0.99	69.34  0.06	76.65  0.02	90.14  0.00	87.29  0.00	94.03  0.00
Notting-Hill	79.53  0.61	48.92  0.00	78.71  0.00	80.86  0.01	70.62  0.00	74.79  0.00	83.58  0.00
NMI
Two-moon	19.79  0.00	15.92  0.00	25.85  0.00	24.10  0.00	69.37  0.00	54.88  0.00	90.22  0.00
Path-based	51.28  0.01	50.92  8.38	57.29  0.00	27.72  2.35	75.63  0.00	75.63  0.00	95.82  0.00
Yale	56.17  1.98	36.27  2.62	61.72  0.92	65.49  0.27	55.40  0.00	59.21  0.00	68.63  0.00
ORL	74.31  1.56	55.26  1.04	82.60  0.72	77.91  0.43	74.07  0.00	79.98  0.00	83.33  0.00
COIL-20	77.83  1.22	55.99  0.47	78.48  0.30	84.83  0.03	94.60  0.00	94.14  0.00	96.42  0.00
Notting-Hill	68.09  1.08	26.94  0.00	64.19  0.00	67.39  0.01	61.80  0.00	71.05  0.00	77.93  0.00