Kernel Parameter Optimization for Kriging Based on Structural Risk Minimization Principle

Table 4

Comparison of the generalization ability between different test functions with nonuniform distribution sampling.

Test functions				RMSE
	CON-Kriging	SRM-Kriging	MLE-Kriging	CON-Kriging	SRM-Kriging	MLE-Kriging
Fun1	Failure	0.955 64	failure	985.97 356	0.049 79	0.564 43
Fun2	1.000 00	0.999 99	0.999 99	0.000 15	0.000 83	0.000 79
Fun3	0.886 67	0.957 91	0.948 37	3.753 33	2.293 57	2.347 32
Fun4	0.474 05	0.983 03	0.999 97	336.163 41	60.180 79	2.715 26
Fun5	0.882 14	0.996 40	0.997 44	354.549 56	61.200 31	61.109 76
Fun6	0.999 90	0.999 98	0.999 99	0.840 81	0.378 94	0.318 47
Fun7	0.424 08	0.747 45	0.677 50	21.787 80	11.903 82	13.896 88
Fun8	Failure	0.338 04	0.009 51	457.409 39	190.136 04	394.725 25
Fun9	0.004 02	0.135 85	0.054 74	2 385.597 16	1 506.804 29	1 889.119 78