`lasso_path` – Lasso with CV-selected alpha (LassoCV).#

Back to family axis | Back to L4 | Browse all options

Operational op under axis family, sub-layer L4_A_model_selection, layer l4. Standalone callable: mf.functions.lasso_path_fit.

Function signature#

mf.functions.lasso_path_fit(
    X: np.ndarray | pd.DataFrame,
    y: np.ndarray | pd.Series,
    *,
    cv: int = 5,
    max_iter: int = 20000,
    random_state: int | None = None,
) -> LassoPathFitResult

Parameters#

name	type	default	constraint	description
`X`	`np.ndarray	pd.DataFrame`	—	—
`y`	`np.ndarray	pd.Series`	—	—
`cv`	`int`	`5`	>=2	Number of cross-validation folds for alpha selection.
`max_iter`	`int`	`20000`	>=1	Maximum coordinate descent iterations per alpha.
`random_state`	`int	None`	`None`	—

Returns#

LassoPathFitResult — frozen dataclass with fit results.

Attribute	Type	Description
`.coef_`	`np.ndarray`	Fitted coefficient vector, shape (n_features,).
`.intercept_`	`float`	Fitted intercept scalar.
`.alpha_selected`	`float`	CV-selected regularisation strength.
`.predict(X)`	`np.ndarray`	Predictions for new data X, shape (n_samples,).
`.summary()`	`str`	Human-readable text table of fit results.

Behavior#

Wraps sklearn’s LassoCV. Picks α automatically from a regularisation path via k-fold CV (params.cv). Equivalent to setting family: lasso, search_algorithm: cv_path.

When to use

When the recipe wants automatic α selection without an explicit search_algorithm.

In recipe context#

Set params.family = "lasso_path" in the relevant layer to activate this op within a recipe:

# Layer L4 recipe fragment
params:
  family: lasso_path

References#

macroforecast design Part 2, L4: ‘forecasting model is the layer where every authoring iteration ends – pick family, tune, repeat.’

lasso_path – Lasso with CV-selected alpha (LassoCV).#