Struct RandomFeaturesRegressor 

pub struct RandomFeaturesRegressor {
    pub feature_map: FourierFeatureMap,
    pub weights: Array1<f64>,
    pub lambda: f64,
    /* private fields */
}

Kernel ridge regression via random Fourier features.

After fitting, prediction is O(D) per test point, making this suitable for large-scale approximate kernel regression.

Example

use scirs2_interpolate::random_features::regressor::RandomFeaturesRegressor;
use scirs2_interpolate::random_features::feature_map::RffKernel;
use scirs2_core::ndarray::{Array1, Array2};

let mut reg = RandomFeaturesRegressor::new(
    RffKernel::Gaussian { length_scale: 1.0 },
    200,   // D random features
    1e-4,  // ridge lambda
    42,    // seed
);
let x = Array2::<f64>::zeros((50, 2));
let y = Array1::<f64>::zeros(50);
reg.fit(&x.view(), &y.view()).expect("fit");
let y_pred = reg.predict(&x.view()).expect("predict");

Fields

feature_map: FourierFeatureMap

Underlying random feature map.

weights: Array1<f64>

Fitted weight vector w (shape [D]), empty until fit() is called.

lambda: f64

Ridge regularization parameter λ.

Implementations

impl RandomFeaturesRegressor

pub fn new(kernel: RffKernel, d_features: usize, lambda: f64, seed: u64) -> Self

Create a new (unfitted) RandomFeaturesRegressor.

Arguments

• kernel — kernel and length-scale for the feature map
• d_features — number of random features D
• lambda — ridge regularization (> 0 recommended for stability)
• seed — RNG seed

Panics

Delegates to FourierFeatureMap::new; panics if d_features == 0. The input dimension d_in is inferred from the first call to fit().

pub fn fit( &mut self, x: &ArrayView2<'_, f64>, y: &ArrayView1<'_, f64>, ) -> Result<(), InterpolateError>

Fit the regressor to training data.

Internally builds Z = feature_map.transform(x) (shape [n, D]) then solves (ZᵀZ + λI)w = Zᵀy via Cholesky decomposition.

Errors

Returns InterpolateError on empty input, shape mismatch, or singular system.

pub fn predict( &self, x: &ArrayView2<'_, f64>, ) -> Result<Array1<f64>, InterpolateError>

Predict at new data points.

Errors

Returns an error if the model has not been fitted yet, or on shape mismatch.

pub fn is_fitted(&self) -> bool

Whether the model has been fitted.

Trait Implementations

impl Clone for RandomFeaturesRegressor

fn clone(&self) -> RandomFeaturesRegressor

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for RandomFeaturesRegressor

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

Formats the value using the given formatter.