Struct TrainableKernel 

pub struct TrainableKernel {
    pub theta: Vec<f32>,
    /* private fields */
}

A trainable quantum embedding kernel with per-feature rotation scales θ.

Fields

theta: Vec<f32>

Trainable per-feature scale parameters.

Implementations

impl TrainableKernel

pub fn new(n_features: usize) -> Self

Construct with all scales initialized to 1.0 (plain angle embedding).

pub fn with_theta(n_features: usize, theta: Vec<f32>) -> QuantumResult<Self>

Construct with explicit initial scales.

Errors

Returns QuantumError::DimensionMismatch if theta.len() != n_features.

pub fn n_features(&self) -> usize

Number of features / qubits.

pub fn kernel(&self, x: &[f32], y: &[f32]) -> QuantumResult<f32>

Kernel value k_θ(x,y) using the current self.theta.

Errors

Propagates embedding errors.

pub fn gram(&self, xs: &[Vec<f32>]) -> QuantumResult<Vec<Vec<f32>>>

Full Gram matrix for xs using the current scales.

Errors

Returns QuantumError::EmptyInput for empty data; propagates embedding errors.

pub fn alignment(&self, xs: &[Vec<f32>], labels: &[f32]) -> QuantumResult<f32>

Kernel-target alignment with the current scales.

Errors

Returns QuantumError::DimensionMismatch if labels.len() != xs.len(), QuantumError::EmptyInput for empty data, and propagates embedding errors.

pub fn train_alignment( &mut self, xs: &[Vec<f32>], labels: &[f32], cfg: &TrainableKernelConfig, ) -> QuantumResult<Vec<f32>>

Train the scales θ by gradient ascent on the kernel-target alignment.

Gradients use the parameter-shift rule on each scale component (the alignment is a smooth function of the scales). Returns the alignment value at every iteration (length iters + 1, including the initial alignment).

Errors

Returns QuantumError::DimensionMismatch on a label/data length mismatch and propagates embedding errors.

Trait Implementations

impl Clone for TrainableKernel

fn clone(&self) -> TrainableKernel

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TrainableKernel

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

Formats the value using the given formatter.