polar_decomposition_newton

scirs2_linalg::decomposition_advanced

Function polar_decomposition_newton 

Source 
pub fn polar_decomposition_newton<A>(
    a: &ArrayView2<'_, A>,
    max_iterations: usize,
    tolerance: A,
) -> LinalgResult<(Array2<A>, Array2<A>)>
where
    A: Float + NumAssign + Debug + Display + ScalarOperand + Sum + Send + Sync + 'static,
Expand description

Iterative refinement for polar decomposition

Uses Newton’s method to iteratively improve the polar decomposition, achieving higher accuracy than the standard SVD-based method.

§Arguments

  • a - Input matrix
  • max_iterations - Maximum number of Newton iterations
  • tolerance - Convergence tolerance

§Returns

  • Tuple (U, P) where U is unitary and P is positive semidefinite