pub fn svd(
matrix: &ArrayView2<'_, f64>,
) -> Result<(Array2<f64>, Array1<f64>, Array2<f64>), String>Expand description
Singular value decomposition is not implemented in this no-SciRS2 fallback (a numerically stable general SVD – e.g. Golub-Kahan bidiagonalization followed by an implicit-shift QR sweep – is well beyond a “naive” reimplementation); honestly report the limitation rather than return a fixed, wrong-sized identity-like result.