#[cfg(feature = "serde-serialize")] use serde::{Deserialize, Serialize}; use crate::allocator::Allocator; use crate::base::{DefaultAllocator, MatrixN, VectorN, U1}; use crate::dimension::Dim; use simba::scalar::ComplexField; /// UDU factorization #[cfg_attr(feature = "serde-serialize", derive(Serialize, Deserialize))] #[derive(Clone, Debug)] pub struct UDU where DefaultAllocator: Allocator + Allocator, { /// The upper triangular matrix resulting from the factorization pub u: MatrixN, /// The diagonal matrix resulting from the factorization pub d: VectorN, } impl Copy for UDU where DefaultAllocator: Allocator + Allocator, VectorN: Copy, MatrixN: Copy, { } impl UDU where DefaultAllocator: Allocator + Allocator, { /// Computes the UDU^T factorization /// The input matrix `p` is assumed to be symmetric and this decomposition will only read the upper-triangular part of `p`. /// Ref.: "Optimal control and estimation-Dover Publications", Robert F. Stengel, (1994) page 360 pub fn new(p: MatrixN) -> Self { let n = p.ncols(); let n_as_dim = D::from_usize(n); let mut d = VectorN::::zeros_generic(n_as_dim, U1); let mut u = MatrixN::::zeros_generic(n_as_dim, n_as_dim); d[n - 1] = p[(n - 1, n - 1)]; u[(n - 1, n - 1)] = N::one(); for j in (0..n - 1).rev() { u[(j, n - 1)] = p[(j, n - 1)] / d[n - 1]; } for j in (0..n - 1).rev() { for k in j + 1..n { d[j] = d[j] + d[k] * u[(j, k)].powi(2); } d[j] = p[(j, j)] - d[j]; for i in (0..=j).rev() { for k in j + 1..n { u[(i, j)] = u[(i, j)] + d[k] * u[(j, k)] * u[(i, k)]; } u[(i, j)] = p[(i, j)] - u[(i, j)]; u[(i, j)] /= d[j]; } u[(j, j)] = N::one(); } Self { u, d } } /// Returns the diagonal elements as a matrix pub fn d_matrix(&self) -> MatrixN { MatrixN::from_diagonal(&self.d) } }