shrinkage::GapSafePenalty

dual_feasible

Source: src/screening.rs

fn dual_feasible(
    &self,
    correlations: &[f64],
    errors: &[f64]
) -> Result<bool, FitError>

Check complete feasibility for all correlations within the error bounds.

Errors

Reject incompatible dimensions or invalid penalty settings.

dual_scaling

Source: src/screening.rs

fn dual_scaling(
    &self,
    _correlations: &[f64],
    _errors: &[f64]
) -> Result<Option<f64>, FitError>

Return an analytic feasible divisor, or request the generic search.

Errors

Reject incompatible dimensions or invalid penalty settings.

has_sparsity

Source: src/screening.rs

fn has_sparsity(&self) -> bool

Whether this penalty has a positive sparsity component.

prox_zeroed

Source: src/screening.rs

fn prox_zeroed(
    &self,
    input: &[f64],
    zeroed: &[bool],
    step: f64,
    output: &mut [f64]
) -> Result<(), FitError>

Evaluate the exact prox with true coordinates constrained to zero.

Errors

Reject invalid parameters, dimensions, steps, or arithmetic.

screen_zeros

Source: src/screening.rs

fn screen_zeros(
    &self,
    correlations: &[f64],
    errors: &[f64],
    zeroed: &mut [bool]
) -> Result<(), FitError>

Add proven zeros using bounds on the optimal correlations.

errors[j] includes the safe radius times the column norm and rounding. Existing true entries must remain fixed.

Errors

Reject incompatible dimensions or invalid penalty settings.