mojolearn.LassoCV

class mojolearn.LassoCV(*, eps=0.001, n_alphas='deprecated', alphas=100, fit_intercept=True, precompute='auto', max_iter=1000, tol=0.0001, copy_X=True, cv=None, verbose=False, n_jobs=None, positive=False, random_state=None, selection='cyclic')

Bases: _LinearRegressorMixin, NumericModeMixin

Lasso with alpha chosen by K-fold cross-validation along a path.

Follows scikit-learn’s LassoCV, solved by coordinate descent: each fold fits the regularization path, the alpha with the lowest mean squared error across folds is chosen, and the model is refit on all rows. cv must be None or an int (unshuffled K-fold); selection='random' raises ValueError.

Parameters:

• eps (float, default 1e-3) – Path length: alpha_min / alpha_max for a generated grid.

• n_alphas (int or ‘deprecated’, default ‘deprecated’) – Grid size, used only when alphas is None; prefer an int alphas.

• alphas (int or sequence of float, default 100) – Number of alphas on a generated grid, or the alphas to try.

• fit_intercept (bool, default True) – Whether to fit an intercept.

• precompute (bool, ‘auto’ or array, default ‘auto’) – Accepted for compatibility; ignored.

• max_iter (int, default 1000) – Maximum number of coordinate-descent passes per alpha.

• tol (float, default 1e-4) – Coordinate-descent stopping tolerance.

• copy_X (bool, default True) – Accepted for compatibility; X is never modified.

• cv (int or None, default None) – Number of folds; None means 5.

• verbose (bool, default False) – Accepted for compatibility; ignored.

• n_jobs (int or None, default None) – Accepted for compatibility; ignored.

• positive (bool, default False) – Restrict the weights to be non-negative.

• random_state (int or None, default None) – Accepted for compatibility; ignored.

• selection ({‘cyclic’}, default ‘cyclic’) – Only ‘cyclic’ is supported.

• numeric_mode (str or None, default None) – Numeric mode for this model ('identical', 'deterministic' or 'fast', which runs on Apple GPUs only); None uses the process default (see mojolearn.set_numeric_mode()).

Variables:

• alpha_ (float) – Selected alpha.

• alphas_ (float32 array of shape (n_alphas,)) – Alphas tried, in decreasing order.

• mse_path_ (float32 array of shape (n_alphas, n_folds)) – Mean squared error of each alpha on each fold.

• coef_ (float32 array of shape (n_features,)) – Weights of the refit model.

• intercept_ (float) – Intercept.

• n_iter_ (int) – Coordinate-descent passes of the refit.

• n_features_in_ (int) – Number of features seen in fit.

Examples

>>> import numpy as np, mojolearn
>>> rng = np.random.default_rng(0)
>>> X = rng.random((1000, 8), dtype=np.float32)
>>> y = 3 * X[:, 0] - 2 * X[:, 1]
>>> model = mojolearn.LassoCV(cv=5).fit(X, y)
>>> model.alpha_

__init__(*, eps=0.001, n_alphas='deprecated', alphas=100, fit_intercept=True, precompute='auto', max_iter=1000, tol=0.0001, copy_X=True, cv=None, verbose=False, n_jobs=None, positive=False, random_state=None, selection='cyclic')

Methods

fit(X, y)	Fit the model and select alpha.
get_params([deep])
numeric_mode_used()	The numeric mode this estimator runs in.
predict(X)	Predict targets for X.
score(X, y)	Coefficient of determination R^2 of predict(X) against y.
set_params(**params)	Set constructor parameters.
vendor_used()	The GPU backend this estimator runs on.

fit(X, y)

Fit the model and select alpha.

Parameters:

• X (array-like of shape (n_samples, n_features)) – Training data, converted to float32; values must be finite.

• y (array-like of shape (n_samples,)) – Targets.

Returns:

self (LassoCV) – The fitted model.

Raises:

ValueError – If selection is not ‘cyclic’, cv is not None or an int in [2, n_samples], or no alpha is given.