Struct QuadraticProbe 

pub struct QuadraticProbe { /* private fields */ }

A separable quadratic probe L(w) = ½ Σ c_i (w_i - center_i)².

Its gradient is g_i = c_i (w_i - center_i) and its Hessian is the diagonal matrix diag(c), so the per-coordinate curvature values are exactly the eigenvalues. This makes it the natural fixture for exercising the condition-number meta-feature: an anisotropic curvature vector produces a known, controllable condition number.

An optional gradient-noise level turns the probe into a stochastic oracle whose per-sample noise is seeded by the sample index, so repeated samples differ yet the whole extraction stays reproducible.

Implementations

impl QuadraticProbe

pub fn new(curvature: Array1<f64>, init: Array1<f64>) -> Result<Self>

Build a bowl with per-coordinate curvature (the Hessian diagonal), centered at the origin, started from init.

pub fn isotropic(dim: usize, curvature: f64, init_value: f64) -> Result<Self>

Build an isotropic dim-dimensional bowl with constant curvature, started from a constant init_value in every coordinate.

pub fn with_center(self, center: Array1<f64>) -> Result<Self>

Override the bowl center (the location of the minimum).

pub fn with_noise(self, noise_std: f64, seed: u64) -> Self

Turn the probe into a stochastic oracle whose gradient samples carry uniform noise of the given standard deviation, seeded by seed.

Trait Implementations

impl Clone for QuadraticProbe

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for QuadraticProbe

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

Formats the value using the given formatter.

impl TaskProbe for QuadraticProbe

fn dimension(&self) -> usize

Dimensionality of the parameter vector.

fn gradient_at(&self, params: &Array1<f64>) -> Array1<f64>

Gradient (or its expectation, for a stochastic oracle) at params.

fn initial_params(&self) -> Array1<f64>

A representative initial parameter vector for the task.

fn gradient_sample( &self, params: &Array1<f64>, sample_index: usize, ) -> Array1<f64>

Draw the sample_index-th (possibly noisy) gradient sample at params.