Struct GaussianProcess 

pub struct GaussianProcess<K: Kernel, P: Prior> {
    pub kernel: K,
    pub prior: P,
    pub noise: f64,
    /* private fields */
}

Gaussian Process for regression

A Gaussian Process defines a distribution over functions. Given training data, it can make predictions with uncertainty estimates.

Fields

kernel: K

Kernel (covariance) function

prior: P

Prior mean function

noise: f64

Noise level (observation noise)

Implementations

impl<K: Kernel, P: Prior> GaussianProcess<K, P>

pub fn new(kernel: K, prior: P, noise: f64) -> Self

Create a new Gaussian Process

pub fn fit(&mut self, x: &Array2<f64>, y: &Array1<f64>) -> StatsResult<()>

Fit the Gaussian Process to training data

pub fn predict(&self, x: &Array2<f64>) -> StatsResult<Array1<f64>>

Predict mean values for new inputs

pub fn predict_with_std( &self, x: &Array2<f64>, ) -> StatsResult<(Array1<f64>, Array1<f64>)>

Predict with mean and standard deviation

pub fn predict_single(&self, x: &ArrayView1<'_, f64>) -> StatsResult<f64>

Predict mean for a single point

pub fn predict_variance_single( &self, x: &ArrayView1<'_, f64>, ) -> StatsResult<f64>

Predict variance for a single point

pub fn log_marginal_likelihood(&self) -> StatsResult<f64>

Compute log marginal likelihood

This is useful for hyperparameter optimization.

pub fn n_train_samples(&self) -> usize

Get number of training samples

Trait Implementations

impl<K: Clone + Kernel, P: Clone + Prior> Clone for GaussianProcess<K, P>

fn clone(&self) -> GaussianProcess<K, P>

Returns a duplicate of the value.

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

Performs copy-assignment from source.