Numeric.LinearAlgebra.SVD.SVDLIBC

Singular value decomposition

svd :: Int -> Matrix Double -> (Matrix Double, Vector Double, Matrix Double)

svd rank a is the sparse SVD of matrix a with the given rank This function handles the conversion to svdlibc's sparse representation.

svd' :: SVDParams -> Int -> Matrix Double -> (Matrix Double, Vector Double, Matrix Double)

Similar to svd but allowing control over the SVDParams.

Sparse SVD

sparsify :: Matrix Double -> CSR

sparseSvd :: Int -> CSR -> (Matrix Double, Vector Double, Matrix Double)

sparseSvd rank a is the sparse SVD of matrix a with the given rank This function handles the conversion to svdlibc's sparse representation, but does not require making the whole matrix dense first

sparseSvd' :: SVDParams -> Int -> CSR -> (Matrix Double, Vector Double, Matrix Double)

Like sparseSvd but providing greater control over the SVDParams used for the computation.

Parameters

data SVDParams

Constructors

SVDParams

Fields

• maxIters :: Maybe Int

  Maximum iteration count

• minEigenvalRange :: (Double, Double)

  Eigenvalues in this range are considered uninteresting

• kappa :: Double

defaultSVDParams :: SVDParams

No iteration limit, exclude eigenvalues in range ((-10^{-30}, +10^{-30}), kappa of 10^{-6}).