Struct RidgePolicy 

pub struct RidgePolicy {
    pub rho_independent: bool,
    pub include_quadratic_penalty: bool,
    pub include_penalty_logdet: bool,
    pub include_laplacehessian: bool,
    pub determinant_mode: RidgeDeterminantMode,
}

Global policy governing how a stabilization ridge participates in objectives.

Fields

rho_independent: bool

Must remain independent of smoothing parameters (rho) for smooth outer derivatives.

include_quadratic_penalty: bool

Include ridge in quadratic penalty term: 0.5 * delta * ||beta||^2.

include_penalty_logdet: bool

Include ridge in penalty determinant term (e.g. log|S_lambda + delta I|).

include_laplacehessian: bool

Include ridge in Hessian used by Laplace term / implicit differentiation.

determinant_mode: RidgeDeterminantMode

Determinant evaluation mode when ridge participates in logdet terms.

Implementations

impl RidgePolicy

pub const fn explicit_stabilization_full() -> RidgePolicy

Default policy used by PIRLS/REML path: treat stabilization ridge as an explicit delta I prior contribution with adaptive logdet evaluation.

pub const fn explicit_stabilization_full_exact() -> RidgePolicy

pub const fn explicit_stabilization_pospart() -> RidgePolicy

Variant used when pseudo-determinants are required for indefinite matrices.

pub const fn solver_only() -> RidgePolicy

Solver-only stabilization: the ridge δI stabilizes the inner linear solve (it bounds the Newton step (H+δI)⁻¹∇) but is excluded from the REML/LAML objective — no ½·δ·‖β‖² quadratic-penalty term, no δ-shift of the penalty log-determinant, no δ-shift of the Laplace Hessian. Use this when a numerical floor is needed purely to keep the linear algebra finite during screening and must NOT bias the smoothing-parameter selection or shrink identified coefficients off the MLE. With every include_* false the optimized objective equals the true penalized REML criterion, so the value surface and its analytic gradient describe the same objective (gam#747/#748).

Trait Implementations

impl Clone for RidgePolicy

fn clone(&self) -> RidgePolicy

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for RidgePolicy

impl Debug for RidgePolicy

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for RidgePolicy

fn deserialize<__D>( __deserializer: __D, ) -> Result<RidgePolicy, <__D as Deserializer<'de>>::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Eq for RidgePolicy

impl PartialEq for RidgePolicy

fn eq(&self, other: &RidgePolicy) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Serialize for RidgePolicy

fn serialize<__S>( &self, __serializer: __S, ) -> Result<<__S as Serializer>::Ok, <__S as Serializer>::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for RidgePolicy