Struct ElasticNetOptions 

pub struct ElasticNetOptions {
    pub lambda: f64,
    pub alpha: f64,
    pub intercept: bool,
    pub standardize: bool,
    pub max_iter: usize,
    pub tol: f64,
    pub penalty_factor: Option<Vec<f64>>,
    pub warm_start: Option<Vec<f64>>,
    pub weights: Option<Vec<f64>>,
    pub coefficient_bounds: Option<Vec<(f64, f64)>>,
}

Options for elastic net fitting.

Configuration options for elastic net regression, which combines L1 and L2 penalties.

Fields

• lambda - Regularization strength (≥ 0, higher = more regularization)
• alpha - Mixing parameter (0 = Ridge, 1 = Lasso, 0.5 = equal mix)
• intercept - Whether to include an intercept term
• standardize - Whether to standardize predictors to unit variance
• max_iter - Maximum coordinate descent iterations
• tol - Convergence tolerance on coefficient changes
• penalty_factor - Optional per-feature penalty multipliers
• warm_start - Optional initial coefficient values for warm starts
• weights - Optional observation weights
• coefficient_bounds - Optional (lower, upper) bounds for each coefficient

Example

let options = ElasticNetOptions {
    lambda: 0.1,
    alpha: 0.5,  // Equal mix of L1 and L2
    intercept: true,
    standardize: true,
    ..Default::default()
};

Fields

lambda: f64

Regularization strength (lambda >= 0)

alpha: f64

Elastic net mixing parameter (0 <= alpha <= 1). alpha=1 is Lasso, alpha=0 is Ridge.

intercept: bool

Whether to include an intercept term

standardize: bool

Whether to standardize predictors

max_iter: usize

Maximum coordinate descent iterations

tol: f64

Convergence tolerance on coefficient changes

penalty_factor: Option<Vec<f64>>

Per-feature penalty factors (optional). If None, all features have penalty factor 1.0.

warm_start: Option<Vec<f64>>

Initial coefficients for warm start (optional). If provided, optimization starts from these values instead of zero. Used for efficient pathwise coordinate descent.

weights: Option<Vec<f64>>

Observation weights (optional). If provided, must have length equal to the number of observations. Weights are normalized to sum to 1 internally.

coefficient_bounds: Option<Vec<(f64, f64)>>

Coefficient bounds: (lower, upper) for each predictor. If None, uses (-inf, +inf) for all coefficients (no bounds).

The bounds vector length must equal the number of predictors (excluding intercept). For each predictor, the coefficient will be clamped to [lower, upper] after each coordinate descent update.

Examples

• Non-negative least squares: Some(vec![(0.0, f64::INFINITY); p])
• Upper bound only: Some(vec![(-f64::INFINITY, 10.0); p])
• Both bounds: Some(vec![(-5.0, 5.0); p])

Notes

• Bounds are applied to coefficients on the ORIGINAL scale, not standardized scale
• The intercept is never bounded
• Each pair must satisfy lower <= upper

Trait Implementations

impl Clone for ElasticNetOptions

fn clone(&self) -> ElasticNetOptions

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ElasticNetOptions

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

Formats the value using the given formatter.

impl Default for ElasticNetOptions

fn default() -> Self

Returns the “default value” for a type.