Struct RandomFeatureInterpolator 

pub struct RandomFeatureInterpolator {
    pub feature_map: RandomFeatureMap,
    pub weights: Vec<f64>,
    pub regularization: f64,
    /* private fields */
}

Kernel ridge regression using random Fourier features.

Solves: (Z^T Z + λ I) w = Z^T y where Z is the random feature matrix. Prediction: f(x*) = z(x*)^T w.

Fields

feature_map: RandomFeatureMap

The underlying random feature map.

weights: Vec<f64>

Solved weight vector, shape [n_features].

regularization: f64

Ridge regularization parameter λ.

Implementations

impl RandomFeatureInterpolator

pub fn new( n_dims: usize, config: RandomFeatureConfig, regularization: f64, ) -> Result<Self, InterpolateError>

Create a new (unfitted) interpolator.

pub fn fit(&mut self, x: &[Vec<f64>], y: &[f64]) -> Result<(), InterpolateError>

Fit to training data (x, y).

Solves (Z^T Z + λ I) w = Z^T y via Cholesky decomposition.

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

Predict at new points.

pub fn kernel_error( &self, x: &[Vec<f64>], true_kernel_fn: impl Fn(&[f64], &[f64]) -> f64, ) -> Result<f64, InterpolateError>

Compute mean absolute error between the random feature kernel approximation and the true kernel evaluated on all pairs in x.

Trait Implementations

impl Clone for RandomFeatureInterpolator

fn clone(&self) -> RandomFeatureInterpolator

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for RandomFeatureInterpolator

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

Formats the value using the given formatter.