Struct GaussianMixtureFit 

pub struct GaussianMixtureFit {
    pub weights: Array1<f64>,
    pub means: Array2<f64>,
    pub covariances: Vec<Array2<f64>>,
    pub k: usize,
    pub d: usize,
    pub n_obs: usize,
    pub loglik: f64,
    pub iterations: usize,
}

A fitted k-component full-covariance Gaussian mixture.

Fields

weights: Array1<f64>

Mixing weights, length k, on the simplex.

means: Array2<f64>

Component means, k × d.

covariances: Vec<Array2<f64>>

Component covariances, k matrices of shape d × d (SPD).

k: usize

Number of mixture components.

d: usize

Data dimension.

n_obs: usize

Number of rows used to fit.

loglik: f64

Maximised total log-likelihood Σ_i log Σ_j w_j N(y_i; μ_j, Σ_j).

iterations: usize

EM iterations taken.

Implementations

impl GaussianMixtureFit

pub fn num_free_parameters(&self) -> usize

Free-parameter count P of a k-component full-covariance mixture in d dimensions: (k − 1) mixing weights on the simplex, k·d mean coordinates, and k · d(d+1)/2 covariance entries. This is the exact quantity that enters the rank-aware normalizer as dim(H) − rank(S).

pub fn per_point_log_density( &self, data: ArrayView2<'_, f64>, ) -> Result<Array1<f64>, String>

Per-observation log predictive density log p(y_i) under the fitted mixture, length n. This is the held-out-density column source for cross-class stacking when the mixture is evaluated on a held-out fold.

pub fn laplace_negative_log_evidence( &self, data: ArrayView2<'_, f64>, ) -> Result<f64, String>

Rank-aware Laplace negative log evidence on the SAME scale as the smooth topology rungs. −V = loglik − ½ log|H| + ½ P log(2π), realised by calling laplace_evidence with residual_objective = −loglik, penalty_log_det = 0, penalty_rank = 0, effective_dim = P, and log|H| the observed empirical-Fisher information at the optimum.

Trait Implementations

impl Clone for GaussianMixtureFit

fn clone(&self) -> GaussianMixtureFit

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for GaussianMixtureFit

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

Formats the value using the given formatter.