Struct TensorSurfaceFit 

pub struct TensorSurfaceFit {
    pub coeffs: Vec<Array2<f64>>,
    pub coeff_covariance: Vec<Array2<f64>>,
    pub residual_cross_cov: Array2<f64>,
    pub unit_covariance: Array2<f64>,
    pub lambda: f64,
    pub edf: f64,
    pub residual_df: f64,
}

A penalized tensor-surface fit over the code sample: the producer of CarveInputs for BOTH binding notions (#993 items 1–2).

coeffs[d] is the fitted M₁ × M₂ coefficient matrix for response dimension d; coeff_covariance[d] is the matching SCALE-INCLUDED posterior covariance of its row-major vec (the mgcv-Vb object wood_smooth_test contracts for); joint_covariance() assembles the cross-dimension covariance for the joint binding test. The fit is evaluated against the SAME empirical code measure the carve centers against — the test and its covariance live on one measure by construction, which is the coherence the production fit’s own Hessian (a different parameterization: tangent frames, not tensor coefficients) cannot offer the carve.

Fields

coeffs: Vec<Array2<f64>>

Per response dimension, M₁ × M₂.

coeff_covariance: Vec<Array2<f64>>

Per response dimension, scale-included Vb of the row-major vec.

residual_cross_cov: Array2<f64>

Scale-included residual cross-covariance between response dimensions (D × D, entries r_dᵀ r_e / (n − edf)). Diagonal entries are the per-dimension scales the Vbs carry.

unit_covariance: Array2<f64>

Scale-FREE coefficient covariance shared by all dimensions (V (Λ+λI)⁻¹ Vᵀ, M₁M₂ × M₁M₂); coeff_covariance[d] is this times residual_cross_cov[d,d].

lambda: f64

REML-selected ridge strength.

edf: f64

Effective degrees of freedom Σ dᵢ/(dᵢ+λ) (per dimension; the design and λ are shared).

residual_df: f64

Residual degrees of freedom n − edf (the denominator d.f. for the Estimated-scale F branch).

Implementations

impl TensorSurfaceFit

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

Joint covariance of the dimension-major stacked coefficient vector [vec(C₀); vec(C₁); …]: with a shared design and shared λ the posterior is the Kronecker product residual_cross_cov ⊗ unit_covariance — index (d·M + i, e·M + j) = S[d,e]·U[i,j]. Feed to CarveInput::joint_coeff_covariance.

Trait Implementations

impl Clone for TensorSurfaceFit

fn clone(&self) -> TensorSurfaceFit

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TensorSurfaceFit

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

Formats the value using the given formatter.