Skip to main content

LinearSolver

apex_solver::linalg

Trait LinearSolver 

Source 
pub trait LinearSolver<M: LinearizationMode> {
    // Required methods
    fn solve_normal_equation(
        &mut self,
        residuals: &Mat<f64>,
        jacobian: &M::Jacobian,
    ) -> LinAlgResult<Mat<f64>>;
    fn solve_augmented_equation(
        &mut self,
        residuals: &Mat<f64>,
        jacobian: &M::Jacobian,
        lambda: f64,
    ) -> LinAlgResult<Mat<f64>>;
    fn get_hessian(&self) -> Option<&M::Hessian>;
    fn get_gradient(&self) -> Option<&Mat<f64>>;

    // Provided methods
    fn compute_covariance_matrix(&mut self) -> Option<&Mat<f64>> { ... }
    fn get_covariance_matrix(&self) -> Option<&Mat<f64>> { ... }
}
Expand description

Unified linear solver interface parameterized by LinearizationMode.

This is the single trait implemented by all linear solvers. When M is a concrete type (e.g., SparseMode), this trait is object-safe and can be used as dyn LinearSolver<SparseMode> or dyn LinearSolver<DenseMode>.

  • Sparse solvers (SparseCholeskySolver, SparseQRSolver, SchurSolverAdapter) implement LinearSolver<SparseMode>.
  • Dense solvers (DenseCholeskySolver, DenseQRSolver) implement LinearSolver<DenseMode>.

Required Methods§

Source

fn solve_normal_equation( &mut self, residuals: &Mat<f64>, jacobian: &M::Jacobian, ) -> LinAlgResult<Mat<f64>>

Solve the normal equations: (J^T · J) · dx = −J^T · r

Source

fn solve_augmented_equation( &mut self, residuals: &Mat<f64>, jacobian: &M::Jacobian, lambda: f64, ) -> LinAlgResult<Mat<f64>>

Solve the augmented equations: (J^T · J + λI) · dx = −J^T · r

Source

fn get_hessian(&self) -> Option<&M::Hessian>

Get the cached Hessian matrix (J^T · J) from the last solve

Source

fn get_gradient(&self) -> Option<&Mat<f64>>

Get the cached gradient vector (J^T · r) from the last solve

Provided Methods§

Source

fn compute_covariance_matrix(&mut self) -> Option<&Mat<f64>>

Compute the covariance matrix (H^{-1}) by inverting the cached Hessian.

Returns None for solvers that do not support covariance estimation (e.g., QR solvers, Schur complement solvers). Only Cholesky-based solvers provide a real implementation.

Source

fn get_covariance_matrix(&self) -> Option<&Mat<f64>>

Get the cached covariance matrix (H^{-1}) computed from the Hessian.

Returns None if covariance has not been computed or is not supported.

Implementors§