Struct linfa_elasticnet::ElasticNetParams

pub struct ElasticNetParams<F> {
    pub penalty: F,
    pub l1_ratio: F,
    pub with_intercept: bool,
    pub max_iterations: u32,
    pub tolerance: F,
}

Linear regression with both L1 and L2 regularization

Configures and minimizes the following objective function: 1 / (2 * n_samples) * ||y - Xw||^2_2 + penalty * l1_ratio * ||w||_1 + 0.5 * penalty * (1 - l1_ratio) * ||w||^2_2

Fields

penalty: F

l1_ratio: F

with_intercept: bool

max_iterations: u32

tolerance: F

Implementations

impl<F: Float> ElasticNetParams<F>

AbsDiffEq + Float + FromPrimitive + ScalarOperand + NumAssignOps> Configure and fit a Elastic Net model

pub fn new() -> ElasticNetParams<F>

Create default elastic net hyper parameters

By default, an intercept will be fitted. To disable fitting an intercept, call .with_intercept(false) before calling .fit().

To additionally normalize the feature matrix before fitting, call fit_intercept_and_normalize() before calling fit(). The feature matrix will not be normalized by default.

pub fn penalty(self, penalty: F) -> Self

Set the overall parameter penalty parameter of the elastic net. Use l1_ratio to configure how the penalty distributed to L1 and L2 regularization.

pub fn l1_ratio(self, l1_ratio: F) -> Self

Set l1_ratio parameter of the elastic net. Controls how the parameter penalty is distributed to L1 and L2 regularization. Setting l1_ratio to 1.0 is equivalent to a “Lasso” penalization, setting it to 0.0 is equivalent to “Ridge” penalization.

Defaults to 0.5 if not set

l1_ratio must be between 0.0 and 1.0.

pub fn with_intercept(self, with_intercept: bool) -> Self

Configure the elastic net model to fit an intercept. Defaults to true if not set.

pub fn tolerance(self, tolerance: F) -> Self

Set the tolerance which is the minimum absolute change in any of the model parameters needed for the parameter optimization to continue.

Defaults to 1e-4 if not set

pub fn max_iterations(self, max_iterations: u32) -> Self

Set the maximum number of iterations for the optimization routine.

Defaults to 1000 if not set

pub fn compute_intercept<'a>(
    &self,
    y: ArrayView1<'a, F>
) -> (F, CowArray<'a, F, Ix1>)

Compute the intercept as the mean of y and center y if an intercept should be used, use 0.0 as intercept and leave y unchanged otherwise.

pub fn validate_params(&self) -> Result<()>

Validate the hyper parameters

This function is called in Self::fit and validates all hyper parameters

Trait Implementations

impl<'a, F, D, T> Fit<'a, ArrayBase<D, Dim<[usize; 2]>>, T> for ElasticNetParams<F> where
    F: Float + AbsDiffEq + Lapack,
    D: Data<Elem = F>,
    T: AsTargets<Elem = F>, 

type Object = Result<ElasticNet<F>>

fn fit(
    &self,
    dataset: &DatasetBase<ArrayBase<D, Ix2>, T>
) -> Result<ElasticNet<F>>

Fit an elastic net model given a feature matrix x and a target variable y.

The feature matrix x must have shape (n_samples, n_features)

The target variable y must have shape (n_samples)

Returns a FittedElasticNet object which contains the fitted parameters and can be used to predict values of the target variable for new feature values.