Struct RationalQuadraticKernel 

pub struct RationalQuadraticKernel {
    pub length_scale: f64,
    pub signal_variance: f64,
    pub alpha: f64,
}

Rational Quadratic kernel.

k(x, x’) = sigma^2 * (1 + ||x - x’||^2 / (2 * alpha * l^2))^(-alpha)

This can be seen as a scale mixture of RBF kernels with different length scales. The parameter alpha controls the relative weighting of large-scale vs small-scale variations.

Fields

length_scale: f64

Length scale

signal_variance: f64

Signal variance

alpha: f64

Shape parameter (alpha); larger alpha => closer to RBF

Implementations

impl RationalQuadraticKernel

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

Trait Implementations

impl Clone for RationalQuadraticKernel

fn clone(&self) -> RationalQuadraticKernel

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for RationalQuadraticKernel

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

Formats the value using the given formatter.

impl Default for RationalQuadraticKernel

fn default() -> Self

Returns the “default value” for a type.

impl SurrogateKernel for RationalQuadraticKernel

fn eval(&self, x1: &ArrayView1<'_, f64>, x2: &ArrayView1<'_, f64>) -> f64

Evaluate the kernel between two input vectors.

fn get_log_params(&self) -> Vec<f64>

Return current hyperparameters as a flat vector (log-scale).

fn set_log_params(&mut self, params: &[f64])

Set hyperparameters from a flat vector (log-scale).

fn clone_box(&self) -> Box<dyn SurrogateKernel>

Clone the kernel into a boxed trait object.

fn name(&self) -> &str

Name of the kernel (for debug display).

fn covariance_matrix(&self, x: &Array2<f64>) -> Array2<f64>

Compute the full covariance matrix for a set of inputs.

fn cross_covariance(&self, x1: &Array2<f64>, x2: &Array2<f64>) -> Array2<f64>

Compute the cross-covariance matrix between two sets of inputs.

fn n_params(&self) -> usize

Number of hyperparameters.