Struct IterativeRefinementLS 

pub struct IterativeRefinementLS {
    pub max_refinement_steps: usize,
    pub inner_tol: f64,
    pub outer_tol: f64,
    pub use_blendenpik: bool,
    pub oversampling: f64,
    pub seed: Option<u64>,
}

Iterative refinement for least-squares with a random preconditioner.

Combines Blendenpik/LSRN preconditioning with iterative refinement steps to achieve high-accuracy solutions even for ill-conditioned problems.

Algorithm:

1. Compute initial solution x₀ using Blendenpik/LSRN
2. Compute residual r = b - A x₀
3. Solve the correction problem: x₁ = x₀ + δ where δ minimizes ‖A δ - r‖
4. Repeat until residual tolerance is met

Fields

max_refinement_steps: usize

Maximum number of refinement steps

inner_tol: f64

Convergence tolerance for each inner solve

outer_tol: f64

Overall convergence tolerance

use_blendenpik: bool

Use Blendenpik (true) or LSRN (false) as inner preconditioner

oversampling: f64

Oversampling factor for the preconditioner

seed: Option<u64>

Random seed

Implementations

impl IterativeRefinementLS

pub fn new() -> Self

Create with default parameters.

pub fn with_max_steps(self, steps: usize) -> Self

Set the maximum number of refinement steps.

pub fn with_inner_tol(self, tol: f64) -> Self

Set inner solver tolerance.

pub fn with_outer_tol(self, tol: f64) -> Self

Set outer convergence tolerance.

pub fn with_lsrn(self) -> Self

Use LSRN instead of Blendenpik.

pub fn with_oversampling(self, alpha: f64) -> Self

Set oversampling factor.

pub fn with_seed(self, seed: u64) -> Self

Set random seed.

pub fn solve( &self, a: &ArrayView2<'_, f64>, b: &ArrayView1<'_, f64>, ) -> LinalgResult<RandomizedPrecondResult>

Solve min ‖A x - b‖ with iterative refinement.

Trait Implementations

impl Clone for IterativeRefinementLS

fn clone(&self) -> IterativeRefinementLS

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for IterativeRefinementLS

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

Formats the value using the given formatter.