NuSVR
#include <Skigen/SVM>
template <typename Scalar = double>
class Skigen::NuSVR(nu=0.5, C=1.0, kernel=Kernel::RBF, degree=3, gamma=0, coef0=0, tol=1e-3, max_iter=1000, random_state=std::nullopt)
Nu-Support Vector Regression with kernels.
In nu-SVR, nu controls the fraction of support vectors and bounds the fraction of points outside the epsilon tube; the tube width epsilon is learned from the data rather than fixed. Skigen fits the dual coefficients with the same sub-gradient solver as SVR, choosing epsilon as the (1 - nu) quantile of the absolute residuals so that roughly a nu fraction of training points lie outside the tube.
Mirrors the dense core of sklearn.svm.NuSVR.
Attributes:
-
nu : Scalar
-
C : Scalar
-
kernel : Kernel
-
epsilon_fitted : Scalar
-
support : const std::vector< Eigen::Index > &
-
n_support : int