Struct VgmmConfig 

pub struct VgmmConfig {
    pub n_components: usize,
    pub prior_concentration: f64,
    pub prior_mean: f64,
    pub prior_precision: f64,
    pub observation_precision: f64,
    pub component_precisions: Option<Vec<f64>>,
    pub max_iterations: usize,
    pub tolerance: f64,
    pub divergence_tolerance: f64,
    pub seed: u64,
}

Configuration for VariationalGaussianMixture.

All fields are validated on VariationalGaussianMixture::new via the private VgmmConfig::validate method. Use the builder helpers to construct a config; the VgmmConfig::new constructor provides sensible research-preview defaults.

Fields

n_components: usize

Number of mixture components K (must be ≥ 1).

prior_concentration: f64

Symmetric Dirichlet prior concentration α₀ > 0 for each component.

prior_mean: f64

Prior mean m₀ for each component’s Gaussian prior.

prior_precision: f64

Prior precision β₀ > 0 for each component’s Gaussian prior.

observation_precision: f64

Shared observation precision τ, used when component_precisions is None. All τ_k are set to this value.

component_precisions: Option<Vec<f64>>

Per-component observation precisions τ_k. When Some, the vector length must equal n_components and every entry must be positive. When None, observation_precision is replicated K times.

max_iterations: usize

Maximum number of VBEM iterations.

tolerance: f64

ELBO absolute-change convergence tolerance.

divergence_tolerance: f64

Maximum permissible ELBO decrease (used for divergence detection).

seed: u64

Seed for the seeded RNG used during mean initialisation.

Implementations

impl VgmmConfig

pub fn new(n_components: usize) -> Self

Construct a config for n_components mixture components with sensible defaults:

Parameter	Default
prior_concentration	1.0
prior_mean	0.0
prior_precision	1e-3
observation_precision	1.0
max_iterations	200
tolerance	1e-6
divergence_tolerance	1e-4
seed	0

pub fn with_prior( self, prior_mean: f64, prior_precision: f64, prior_concentration: f64, ) -> Self

Set the Gaussian prior hyperparameters for all component means.

• prior_mean — prior mean m₀
• prior_precision — prior precision β₀ (must be > 0)
• prior_concentration — symmetric Dirichlet concentration α₀ (must be > 0)

pub fn with_observation_precision(self, tau: f64) -> Self

Set the shared observation precision τ applied to all components.

pub fn with_component_precisions(self, taus: Vec<f64>) -> Result<Self>

Set per-component observation precisions τ_k. The vector length must equal n_components and every entry must be strictly positive; this is validated immediately.

pub fn with_limits(self, max_iterations: usize, tolerance: f64) -> Self

Set iteration budget and ELBO tolerance.

pub fn with_seed(self, seed: u64) -> Self

Set the seed for the seeded RNG used during initialisation.

Trait Implementations

impl Clone for VgmmConfig

fn clone(&self) -> VgmmConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for VgmmConfig

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

Formats the value using the given formatter.