An Optimized Neural Network Classification Method Based on Kernel Holistic Learning and Division

Table 3

Performance comparison on benchmark large sample datasets.

Datasets	Methods	Training time (s)	Testing		No. of kernels
			Svmguide1	ELM		0.15	90.52	90.36	200
KHLD + ELM	32.53	89.81		89.73	442,200
HSARBF-ELM	387.52	91.76		91.34	442,160
KHLD + HSARBF-ELM	423.64	91.24		90.81
CO-BP	18.54	90.41		90.13	30
KHLD + CO-BP	55.37	89.38		89.22	442,30
FOGDM-RBF	28.53	92.34		92.18	70
KHLD + FOGDM-RBF	65.70	91.82		91.56	442,70
Wilt	ELM	0.06	62.63	60.40	100
	KHLD + ELM	37.91	61.82	59.84	173,100
	HSARBF-ELM	358.61	64.73	63.93	173,80
	KHLD + HSARBF-ELM	387.42	64.17	63.49
	CO-BP	38.52	60.28	59.52	30
	KHLD + CO-BP	78.65	59.56	58.72	173,30
	FOGDM-RBF	24.73	62.85	62.37	60
	KHLD + FOGDM-RBF	63.85	62.39	62.13	173,60