Function make_lambda_path 

pub fn make_lambda_path(
    x: &Matrix,
    y: &[f64],
    options: &LambdaPathOptions,
    penalty_factor: Option<&[f64]>,
    intercept_col: Option<usize>,
) -> Vec<f64>

Generates a lambda path.

lambda[0] = inf                  // Large value (all coefficients zero)
lambda[1] = lambda_decay_factor * lambda_max           // First real lambda
lambda[k] = lambda[k-1] * lambda_decay_factor    // Geometric decay

Where:

• lambda_max = max(|X^T y|) / max(alpha, 1e-3): Theoretical lambda_max
• lambda_decay_factor = max(lambda_min_ratio, eps)^(1/(nlambda-1)): Geometric decay factor
• eps = 1.0e-6: Minimum lambda_min_ratio value

Arguments

• x - Standardized design matrix (n × p), columns are unit norm
• y - Standardized response vector (||y|| = 1)
• options - Lambda path generation options
• penalty_factor - Optional per-feature penalty factors
• intercept_col - Index of intercept column (typically 0)

Returns

A vector of lambda values in decreasing order.

First Lambda (LAMBDA_EFFECTIVE_INFINITY)

The first lambda value is set to infinity, which effectively produces all-zero coefficients. This matches R’s cursed behavior. PROOF OF CONCEPT/R EQUIVALENCE: We may want to make this optional in future versions.

Default lambda_min_ratio

Following glmnet:

• If n >= p: lambda_min_ratio = 0.0001
• If n < p: lambda_min_ratio = 0.01