Table 3.

Classification results with their corresponding RMSE and computational time for the automatically selected features (d refers to the intrinsic dimension, and f is the dimension of the NCA-based selected feature) using two values of NCA tolerance and three compressive sampling rates.

NCA Tolerance Value	CS Sampling Rate $\left(\alpha \right)$	d	f	MLR Classifier
				Training Accuracy (%)	Training Time (s)	Testing Accuracy (%)	Testing Time (s)
0.01	0.1	28	8	99.8 ± 0.3	5.34 ± 1.7	99.5 ± 0.6	0.015 ± 0.002
	0.2	40	10	99.9 ± 0.1	4.6 ± 2.3	99.7 ± 0.3	0.003 ± 0.00
	0.3	26	8	100 ± 0.0	3.3 ± 0.5	99.9 ± 0.1	0.003 ± 0.001
0.02	0.1	62	18	99.9 ± 0.2	3.37 ± 0.8	99.7 ± 0.3	0.003 ± 0.001
	0.2	55	14	99.9 ± 0.1	3.55 ± 0.9	99.8 ± 0.2	0.003 ± 0.001
	0.3	33	11	100 ± 0.0	4.6 ± 2.0	100 ± 0.0	0.004 ± 0.003