An Extension of the Quadratic Error Function for Learning Imprecise Data in Multivariate Nonlinear Regression
Table 1
Comparison of statistical performances.
Error function
Constrained on the errors
Errors are centered, reduced, and uncorrelated
0.599
3.586
0.658
3.365
0.857
1.011
0.819
4.646
0.452
2.831
0.783
0.845
0.599
3.586
0.658
3.365
0.857
1.011
0.819
4.646
0.452
2.831
0.783
0.845
0.643
2.434
0.787
2.105
0.933
0.415
0.902
2.917
0.929
1.783
0.945
0.326
Errors are centered, reduced, and correlated
0.705
2.733
0.772
1.758
0.821
0.681
0.493
1.734
0.590
0.718
0.630
0.429
0.798
2.069
0.832
1.334
0.899
0.511
0.588
1.240
0.667
0.526
0.755
0.297
0.805
1.540
0.901
1.258
0.912
0.426
0.693
1.011
0.704
0.333
0.919
0.134
= classical generalized least squares; = extension of generalized least squares proposed by1; = new extension of generalized least squares proposed by us; = coefficient of determination; = mean absolute percentage error.