Struct GaussianMixtureModel 

pub struct GaussianMixtureModel<F> {
    pub n_components: usize,
    pub config: GMMConfig,
    pub parameters: Option<GMMParameters<F>>,
    pub convergence_history: Vec<F>,
    /* private fields */
}

Gaussian Mixture Model with EM algorithm

Fields

n_components: usize

Number of components

config: GMMConfig

Configuration

parameters: Option<GMMParameters<F>>

Fitted parameters

convergence_history: Vec<F>

Convergence history

Implementations

impl<F: GmmFloat> GaussianMixtureModel<F>

pub fn new(n_components: usize, config: GMMConfig) -> StatsResult<Self>

Create new Gaussian Mixture Model

pub fn fit( &mut self, data: &ArrayView2<'_, F>, ) -> StatsResult<&GMMParameters<F>>

Fit GMM to data using EM algorithm

pub fn predict(&self, data: &ArrayView2<'_, F>) -> StatsResult<Array1<usize>>

Predict cluster assignments (hard assignment: argmax of responsibilities)

pub fn predict_proba(&self, data: &ArrayView2<'_, F>) -> StatsResult<Array2<F>>

Predict soft cluster assignment (responsibility matrix)

pub fn score(&self, data: &ArrayView2<'_, F>) -> StatsResult<F>

Average log-likelihood per sample

pub fn score_samples(&self, data: &ArrayView2<'_, F>) -> StatsResult<Array1<F>>

Per-sample log-likelihood

pub fn sample(&self, n: usize, seed: Option<u64>) -> StatsResult<Array2<F>>

Generate random samples from the fitted mixture model

pub fn bic(&self, _data: &ArrayView2<'_, F>) -> StatsResult<F>

Bayesian Information Criterion for the fitted model

pub fn aic(&self, _data: &ArrayView2<'_, F>) -> StatsResult<F>

Akaike Information Criterion for the fitted model

pub fn n_parameters(&self) -> StatsResult<usize>

Number of free parameters in the model