pub const SVM_EPS_SVR: i32 = 103;
-Support Vector Regression. The distance between feature vectors from the training set and the fitting hyper-plane must be less than p. For outliers the penalty multiplier C is used.