This computes a thin SVD using a symmetric eigendecomposition of A.T @ A
or A @ A.T. It is intended as a baseline implementation; it is less
numerically stable than a dedicated SVD routine for ill-conditioned inputs.
With computeUv: true, returns [u, s, vh] such that
a ~= u @ diag(s) @ vh. Only fullMatrices: false is supported for
non-square matrices.
Singular value decomposition of real matrices.
This computes a thin SVD using a symmetric eigendecomposition of
A.T @ AorA @ A.T. It is intended as a baseline implementation; it is less numerically stable than a dedicated SVD routine for ill-conditioned inputs.With
computeUv: true, returns[u, s, vh]such thata ~= u @ diag(s) @ vh. OnlyfullMatrices: falseis supported for non-square matrices.