Similarity Measure Learning in Closed-Form Solution for Image Classification

Table 4

Classification performance comparison on the MNIST database.

Method	50 Train		100 Train		150 Train		200 Train		250 Train		300 Train
	Error (%)	d	Error (%)	d	Error (%)	d	Error (%)	d	Error (%)	d	Error (%)	d
PCA	16.1±0.72	499	10.9±0.46	517	9.2±0.48	561	7.8±0.33	578	7.0±0.35	603	7.0±0.28	610
RS-2DPCA	11.2±0.42	18	7.3±0.26	18	4.5±0.21	18	3.8±0.19	18	3.3±0.20	18	2.0±0.19	18
LDA	12.4±0.53	9	9.2±0.39	9	8.6±0.27	9	7.0±0.22	9	5.4±0.24	9	4.6±0.17	9
LPP	10.7±0.47	56	6.7±0.21	51	4.8±0.25	43	3.5±0.17	58	4.5±0.14	69	1.9±0.20	73
MFA	10.5±0.42	114	7.1±0.27	108	4.0±0.23	121	3.7±0.19	98	3.0±0.17	105	1.8±0.14	94
CDA	11.6±0.38	49	6.2±0.29	63	3.4±0.26	70	3.3±0.24	62	2.9±0.22	54	2.2±0.23	68
CEA	12.1±0.43	31	5.9±0.35	52	3.1±0.16	62	3.8±0.18	60	2.7±0.17	83	1.6±0.19	76
ISM-GE	10.8±0.39	92	6.1±0.32	91	3.3±0.20	87	2.6±0.17	94	1.9±0.15	90	1.8±0.17	91
MSE	11.5±0.41	83	6.4±0.28	85	3.7±0.17	82	3.1±0.22	87	2.7±0.21	84	2.3±0.19	83
CSML	9.4 ± 0.31	62	5.3 ± 0.26	79	3.6 ± 0.18	85	1.9 ± 0.16	83	1.3 ± 0.18	83	1.2 ± 0.13	89
KCSML	8.7 ± 0.36	74	4.7 ± 0.34	82	2.4 ± 0.16	85	2.2 ± 0.17	89	1.5 ± 0.12	93	1.1 ± 0.16	90