Struct ChebyshevPreconditioner 

pub struct ChebyshevPreconditioner { /* private fields */ }

Chebyshev polynomial preconditioner with automatic eigenvalue estimation.

Constructs p(A) ~ A^{-1} where p is a degree-d Chebyshev polynomial optimally scaled to the spectrum [lambda_min, lambda_max].

Application proceeds via the three-term Chebyshev recurrence, requiring degree sparse matrix-vector products per application.

Implementations

impl ChebyshevPreconditioner

pub fn from_csr( csr: &CsrMatrix<f64>, config: &PolyPrecondConfig, ) -> SparseResult<Self>

Build a Chebyshev preconditioner from a CSR matrix.

Eigenvalue bounds are estimated automatically using a few Lanczos steps.

Arguments

• csr - The SPD sparse matrix in CSR format.
• config - Configuration parameters.

pub fn with_bounds( csr: &CsrMatrix<f64>, degree: usize, lambda_min: f64, lambda_max: f64, ) -> SparseResult<Self>

Build a Chebyshev preconditioner with explicit eigenvalue bounds.

Arguments

• csr - The SPD sparse matrix in CSR format.
• degree - Polynomial degree (5-20 recommended).
• lambda_min - Lower bound on the eigenvalue spectrum.
• lambda_max - Upper bound on the eigenvalue spectrum.

pub fn apply(&self, r: &[f64]) -> SparseResult<Vec<f64>>

Apply the Chebyshev polynomial preconditioner: z = p(A) * r.

Uses the Chebyshev iteration three-term recurrence. This computes an approximation to A^{-1} r without constructing any matrix.

pub fn eigenvalue_bounds(&self) -> (f64, f64)

Return the estimated eigenvalue bounds.

pub fn degree(&self) -> usize

Return the polynomial degree.

pub fn size(&self) -> usize

Return the matrix dimension.