An ECoG-Based Binary Classification of BCI Using Optimized Extreme Learning Machine

Table 2

Objective classification performance of ELM.

Parameters			Evaluation metrics
Activation function	Sample point	Hidden layer nodes	Classification accuracy (%)	Average training time (s)	Average testing time (s)
sin	1500	41	75.64	0.0009	<0.0001
	2000	58	80.77	0.0043	0.0012
	2500	57	76.92	0.0055	0.0012
sig	1500	34	78.21	0.0040	0.0034
	2000	49	79.49	0.0018	0.0018
	2500	32	80.77	0.0006	0.0006
hardlim	1500	25	67.95	0.0018	0.0012
	2000	50	78.21	0.0015	0.0031
	2500	18	74.32	0.0012	0.0012
tribas	1500	40	67.95	0.0015	0.0049
	2000	57	71.79	0.0015	<0.0001
	2500	53	69.23	0.0018	0.0012
radbas	1500	43	71.79	0.0021	0.0006
	2000	58	69.23	0.0043	0.0006
	2500	50	75.64	0.0012	0.0012