Struct Lasso 

pub struct Lasso { /* private fields */ }

Lasso regression with an unpenalised intercept, fitted by cyclic coordinate descent.

Minimises 1/(2n)·‖y − b − Xw‖² + α‖w‖₁, the same objective as scikit-learn’s Lasso(alpha=α). The classic DML learner for high-dimensional confounders.

Implementations

impl Lasso

pub fn new(alpha: f64) -> Self

pub fn tol(self, tol: f64) -> Self

Stopping rule of scikit-learn’s Lasso(tol=tol) (default 1e-4): once no coefficient moves by more than tol · max|w| in a sweep, stop if the duality gap, which bounds the distance to the optimal objective, is at most tol · ‖y − ȳ‖² / n. With α = 0 only the first condition applies.

pub fn max_iter(self, max_iter: usize) -> Self

pub fn coef(&self) -> Option<&Array1<f64>>

pub fn intercept(&self) -> f64

Trait Implementations

impl Clone for Lasso

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for Lasso

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Regressor for Lasso

fn fit_weighted( &mut self, x: ArrayView2<'_, f64>, y: ArrayView1<'_, f64>, w: ArrayView1<'_, f64>, ) -> Result<()>

Minimises 1/(2n)·Σ wᵢ rᵢ² + α‖β‖₁ with weights rescaled to sum to n, as scikit-learn’s Lasso.fit(sample_weight=w).

fn fit(&mut self, x: ArrayView2<'_, f64>, y: ArrayView1<'_, f64>) -> Result<()>

fn predict(&self, x: ArrayView2<'_, f64>) -> Result<Array1<f64>>

fn fresh(&self) -> Box<dyn Regressor>