ridge - direct closed-form Ridge regression

Group: Regularised · ABI: n4m_ridge_fit

Description

ridge fits the native closed-form multi-output Ridge model:

beta = (Xc'Xc + alpha I)^-1 Xc'Yc
intercept = y_mean - x_mean beta

The solver is selected by shape: primal augmented-QR for p <= n, dual Gram-on-samples for p > n. This is distinct from ridge_pls, which is ridge-augmented SIMPLS.

Python Usage

import n4m

res = n4m.ridge(X, y, alpha=0.1)
y_hat = res["predictions"]
coef = res["coefficients"]
intercept = res["intercept"]

Returned keys include coefficients, intercept, x_mean, x_scale, y_mean, predictions, scalar rmse, and scalar lambda.

Reusable sklearn-style wrapper:

from n4m.sklearn import NativeRidgeRegressor

model = NativeRidgeRegressor(alpha=0.1).fit(X, y)
y_hat = model.predict(X_test)

C ABI

n4m_ridge_fit(ctx, cfg, &x_view, &y_view,
              &alpha, 1, &result);