SVR

#include <Skigen/SVM>

template <typename Scalar = double>
class Skigen::SVR(C=1.0, kernel=Kernel::RBF, degree=3, gamma=0, coef0=0, epsilon=0.1, tol=1e-3, max_iter=1000, random_state=std::nullopt)

Epsilon-Support Vector Regression with kernels.

Mirrors sklearn.svm.SVR.

Solves the $\epsilon$-insensitive primal in feature space using a sub-gradient SGD over the kernel-mapped dual coefficients. The solver is simpler than libsvm's full SMO-for-regression — it converges to the same minimiser at the optimum but the iterate trace differs.

---

---

Attributes:

• C : Scalar

• epsilon : Scalar

• kernel : Kernel

• support : const std::vector< Eigen::Index > &

• n_support : int

---

Methods

SKIGEN_PARAMS()

---

fit(X, y)

Fit from a sparse design matrix (densifies internally).

---

predict(X)

---

score(X, y)

---