An Improved Kernel Based Extreme Learning Machine for Robot Execution Failures

Table 5

Comparison of performance by AKELM and KELM learning algorithms for the classification problems.

Algorithms with different kernel functions	Wine			Diabetes
	Training accuracy	Testing accuracy	Training time; (seconds)	Training accuracy	Testing accuracy	Training time; (seconds)
KELM (parameters = 1, Gaussian)	100%	100%	0.0277	84.38%	77.08%	0.1394
KELM (parameters = 1, tangent)	51.33%	50%	0.0067	73.78%	73.44%	0.1326
KELM (parameters = 1, wavelet)	100%	100%	0.0070	86.81%	76.56%	0.1347
KELM (parameters = 10, Gaussian)	100%	100%	0.0083	78.99%	79.17%	0.0919
KELM (parameters = 10, tangent)	39.33%	42.86%	0.0023	65.80%	65.63%	0.0904
KELM (parameters = 10, wavelet)	100%	96.43%	0.0061	80.03%	77.08%	0.1361
AKELM (Gaussian)	100%	100%	17.8594	90.45%	80.21%	260.7031
AKELM (tangent)	97.33%	100%	13.9375	73.26%	79.17%	313.8750
AKELM (wavelet)	100%	100%	16	89.06%	79.69%	335.5469