An Effective Clustering Algorithm Using Adaptive Neighborhood and Border Peeling Method

Table 2

Comparison with BP clustering algorithm on MNIST dataset samples.

	D1 (r = 120, classes = 7)				D2 (r = 130, classes = 6)				D3 (r = 140, classes = 7)
	ARI	AMI	Det#	K	ARI	AMI	Det#	K	ARI	AMI	Det#	K
BP	0.977	0.962	7.3	20	0.930	0.892	6.3	20	0.860	0.886	7.7	20
SANBP	0.980	0.865	7.0	31	0.921	0.865	6.0	30	0.920	0.888	7.1	34

All the bold values identify the best result.