Struct KernelSelector 

pub struct KernelSelector { /* private fields */ }

Kernel selector for choosing and comparing kernels.

Implementations

impl KernelSelector

pub fn new() -> Self

Create a new kernel selector.

pub fn with_regularization(self, reg: f64) -> Self

Set the regularization parameter.

pub fn kernel_target_alignment<K: Kernel + ?Sized>( &self, kernel: &K, data: &[Vec<f64>], targets: &[f64], ) -> Result<f64>

Compute Kernel Target Alignment (KTA).

KTA measures how well a kernel aligns with the target labels. Higher values indicate better alignment.

Formula: KTA = <K, yy^T>_F / (||K||_F * ||yy^T||_F)

Arguments

• kernel - The kernel to evaluate
• data - Input data points
• targets - Target values (for regression) or labels (for classification)

pub fn centered_kernel_target_alignment<K: Kernel + ?Sized>( &self, kernel: &K, data: &[Vec<f64>], targets: &[f64], ) -> Result<f64>

Compute centered Kernel Target Alignment.

Centered KTA is more robust and accounts for the mean of the kernel matrix.

pub fn compare_kernels_kta( &self, kernels: &[(&str, &dyn Kernel)], data: &[Vec<f64>], targets: &[f64], ) -> Result<KernelComparison>

Compare multiple kernels using KTA.

Returns a comparison with the best kernel identified.

pub fn loo_error_estimate<K: Kernel + ?Sized>( &self, kernel: &K, data: &[Vec<f64>], targets: &[f64], ) -> Result<f64>

Evaluate kernel quality using Leave-One-Out (LOO) error estimate.

For GP regression, this provides an efficient estimate of generalization error. Lower values indicate better performance.

Note: This is an approximation based on the kernel matrix.

pub fn k_fold_cv<K: Kernel + ?Sized>( &self, kernel: &K, data: &[Vec<f64>], targets: &[f64], config: &KFoldConfig, ) -> Result<CrossValidationResult>

Perform K-fold cross-validation for kernel evaluation.

This evaluates a kernel by training on K-1 folds and testing on the remaining fold. The score returned is based on kernel alignment with targets.

pub fn grid_search_rbf_gamma( &self, data: &[Vec<f64>], targets: &[f64], gammas: &[f64], ) -> Result<GammaSearchResult>

Find the best gamma for RBF kernel using grid search.

Searches over a logarithmic grid of gamma values.

Trait Implementations

impl Clone for KernelSelector

fn clone(&self) -> KernelSelector

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for KernelSelector

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

Formats the value using the given formatter.

impl Default for KernelSelector

fn default() -> KernelSelector

Returns the “default value” for a type.