Struct GaussianProcess 

pub struct GaussianProcess {
    pub length_scale: f64,
    pub signal_variance: f64,
    pub noise_variance: f64,
    /* private fields */
}

Gaussian Process regression with an RBF kernel.

Hyperparameters are settable; the kernel is k(x, x') = σ_f² · exp(−||x−x'||² / 2ℓ²) with observation noise σ_n.

Fields

length_scale: f64

Length scale ℓ.

signal_variance: f64

Signal variance σ_f².

noise_variance: f64

Observation noise variance σ_n².

Implementations

impl GaussianProcess

pub const fn new( length_scale: f64, signal_variance: f64, noise_variance: f64, ) -> Self

Create a GP with the given kernel hyperparameters.

pub fn n_samples(&self) -> usize

Number of training samples.

pub fn add_sample(&mut self, x: Vec<f64>, y: f64)

Add a training sample (x, y).

pub fn set_hyperparameters( &mut self, length_scale: f64, signal_variance: f64, noise_variance: f64, )

Set kernel hyperparameters (all clamped to positive).

pub fn log_marginal_likelihood(&self) -> Result<f64, String>

Log marginal likelihood of the training data under the current hyperparameters: −½ yᵀK⁻¹y − ½ log|K| − ½ n log 2π.

Requires a successful fit. Used for hyperparameter optimization — higher is better.

pub fn fit_hyperparameters( &mut self, n_initial: usize, iterations: usize, n_candidates: usize, seed: u64, ) -> Result<(), String>

Optimize the kernel hyperparameters (length scale, signal variance, noise variance) by maximizing the log marginal likelihood.

Uses the crate’s own BayesianOptimizer over log-space hyperparameters (dogfooding). iterations GP refits after the initial random search; n_candidates EI candidates per iteration.

pub fn fit(&mut self) -> Result<(), String>

Fit the GP: compute L = cholesky(K + σ_n² I) and α = K⁻¹ y.

Returns an error if fewer than 2 samples are present.

pub fn predict(&self, x: &[f64]) -> Result<(f64, f64), String>

Predict mean and variance at a query point.

Returns (mean, variance); variance includes observation noise. Errors if not fitted.

pub const fn min_eigenvalue(&self) -> f64

Minimum Cholesky eigenvalue during fit (diagnostic).

Trait Implementations

impl Clone for GaussianProcess

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for GaussianProcess

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

Formats the value using the given formatter.

impl Default for GaussianProcess

fn default() -> Self

Returns the “default value” for a type.