Struct Pce 

pub struct Pce {
    pub indices: Vec<Vec<usize>>,
    pub coefficients: Vec<f64>,
    pub total_variance: f64,
    pub sobol_first_order: Vec<f64>,
    pub sobol_total: Vec<f64>,
    pub r_squared: f64,
}

Polynomial chaos expansion — a Hermite-polynomial surrogate with analytic Sobol’ sensitivity indices.

Fields

indices: Vec<Vec<usize>>

Multi-indices of the retained basis terms.

coefficients: Vec<f64>

Regression coefficients per term.

total_variance: f64

Total variance of the surrogate (Σ c²).

sobol_first_order: Vec<f64>

Sobol’ main-effect indices per input dimension (first-order).

sobol_total: Vec<f64>

Sobol’ total-effect indices per input dimension.

r_squared: f64

R² of the surrogate fit (determination coefficient).

Implementations

impl Pce

pub fn fit( x: &[Vec<f64>], y: &[f64], bounds: &[(f64, f64)], max_degree: usize, ) -> Self

Fit a PCE to the given samples.

x inputs must be normalized to the box bounds (each column in [lo, hi]); standardization to N(0,1) happens internally via ξ = 2·(x − lo)/(hi − lo) − 1 (an affine map — acceptable for PCE on bounded boxes).