Struct MarginalSlopeCovariance 

pub struct MarginalSlopeCovariance { /* private fields */ }

Immutable, validated covariance geometry for the physical log-slope vector.

Admission is the single validation boundary. Diagonal covariance entries remain the sole authority for their quadratic forms. Full covariances cache an eigensquare-root factor so subsequent quadratic forms are exact sums of squares; runtime code never repeats an eigendecomposition or applies a negative-value tolerance. The exact 1ᵀΣ1 shared-slope geometry is cached at the same boundary.

Implementations

impl MarginalSlopeCovariance

pub fn diagonal(covariance: Array1<f64>) -> Result<Self, String>

pub fn full(covariance: Array2<f64>) -> Result<Self, String>

pub fn low_rank(factor: Array2<f64>) -> Result<Self, String>

pub fn to_dense(&self) -> Array2<f64>

pub fn shape(&self) -> MarginalSlopeCovarianceShape

pub fn dim(&self) -> usize

pub fn ones_quadratic_form(&self) -> f64

pub fn quadratic_form(&self, vector: &[f64]) -> Result<f64, String>

Trait Implementations

impl Clone for MarginalSlopeCovariance

fn clone(&self) -> MarginalSlopeCovariance

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for MarginalSlopeCovariance

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

Formats the value using the given formatter.

impl From<MarginalSlopeCovariance> for ScoreCovarianceField

fn from(pooled: MarginalSlopeCovariance) -> Self

Converts to this type from the input type.

impl PartialEq for MarginalSlopeCovariance

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

Equality operator ==.

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

Inequality operator !=.

impl SymmetricQuadraticCoefficients for MarginalSlopeCovariance

fn dimension(&self) -> usize

Input-space dimension of the symmetric operator.

fn multiply(&self, input: &[f64], output: &mut [f64])

Compute output = A * input without materializing A.

fn coefficient(&self, row: usize, column: usize) -> f64

One symmetric coefficient A[row, column].

fn visit_upper_triangle( &self, direction: &mut [f64], projected: &mut [f64], visit: impl FnMut(usize, usize, f64), )

Visit the operator’s input-space upper triangle without materializing a dense matrix. The default probes one basis direction at a time through Self::multiply, preserving matrix-free operators. Representations with direct structure should override this hook so a packed curvature lowering is O(K), O(K²), or O(K²R) as appropriate instead of performing K full operator applications.

fn quadratic_value<T, F>(&self, input: &[T], value: F) -> f64

where F: Fn(&T) -> f64,

Evaluate the primal quadratic form without materializing the input as a separate f64 vector. Structured operators should override this to preserve their representation’s natural complexity (for example O(KR) for a K-by-R low-rank factor).