We choose ?-SVR for nonlinear approximations for your corrected coupling features, and that is much not as likely to subject to overfitting difficulties.?Also, the fitting processes of People features are independent from just one dimension to a different, that's why the fitting results of 1 dimension does not affect that of A further Proportions.?