LinearSVC

#include <Skigen/SVM>

template <typename Scalar = double>
class Skigen::LinearSVC(C=1.0, loss=Loss::SquaredHinge, tol=1e-4, max_iter=1000, fit_intercept=true, random_state=std::nullopt)

Linear SVC — primal sub-gradient SGD on the squared-hinge L2 regularised objective.

Mirrors the binary case of sklearn.svm.LinearSVC; multiclass uses one-vs-rest (default multi_class="ovr").

Solves $\min_w \tfrac{1}{2}\|w\|^2 + C \sum_i \max(0, 1 - y_i (w \cdot x_i + b))^2$ (loss="squared_hinge", the sklearn default) or the analogous hinge loss when loss="hinge".

---

---

Attributes:

• C : Scalar

• loss : Loss

• classes : const Eigen::VectorXi

• n_classes : int

• coef : MatrixType

• intercept : VectorType

---

Methods

SKIGEN_PARAMS()

---

fit(X, y)

Fit from a sparse design matrix (densifies internally).

---

predict(X)

---

predict(X)

Predict from a sparse design matrix (densifies internally).

---

decision_function(X)

Raw decision function. Shape is (n_samples,) for binary problems wrapped as a 1-column matrix; (n_samples, n_classes) for multiclass.

---

fit(X, y)

---

predict(X)

---