UMFFactor¶

class sksparse.umfpack.UMFFactor[source]¶

The main object used for creating and using an LU factorization.

The constructor computes the symbolic analysis of a sparse matrix \(A\) and determines a fill-reducing ordering such that:

\[L U = P R A Q.\]

The numeric factorization is not computed until factorize() is called.

Note

Note that the use of the scale factor R differs between KLU and UMFPACK:

\[\begin{split}L U &= P R_{\mathrm{umf}} A Q \quad &&\text{(UMFPACK)}, \\ L U + F &= R_{\mathrm{klu}} P A Q \quad &&\text{(KLU)}.\end{split}\]

They are related by \(R_{\mathrm{klu}} = P R_{\mathrm{umf}} P^{\top}\).

is_numeric[source]¶

Whether the numeric factorization has been computed.

Type:: bool

lnz, unz

Number of nonzeros in \(L\) and \(U\), respectively.

Type:: int

nnz[source]¶

Total number of nonzeros in \(L\) and \(U\).

Type:: int

n_row, n_col

Number of rows and columns in the input matrix.

Type:: int

nz_udiag[source]¶

Number of nonzeros on the diagonal of \(U\).

Type:: int

dtype[source]¶

The data type of the matrix entries (float64 or complex128).

Type:: numpy.dtype

itype[source]¶

The integer type used for indexing (int32 or int64).

Type:: numpy.dtype

L[source]¶

The \(L\) factor as a sparse CSR matrix.

Type:: scipy.sparse.csr_array

U[source]¶

The \(U\) factor as a sparse CSC matrix.

Type:: scipy.sparse.csc_array

perm_r, perm_c

The row and column permutation arrays, \(P\) and \(Q\).

Type:: numpy.ndarray

R[source]¶

The row scaling diagonal matrix as a 1D array.

Type:: numpy.ndarray

info[source]¶

An object containing information about the factorization.

Type:: UMFInfo

control[source]¶

An object containing settings for the factorization.

Type:: UMFControl

See also

UMFControl, umf_factor, umf_solve

Notes

This object is an interface to the SuiteSparse UMFPACK library [1].

Added in version 0.5.0.

References

Methods

`__init__`	Compute the symbolic analysis.
`copy`	Return a deep copy of the current UMFFactor object.
`factorize`	Compute the numeric factorization of a sparse matrix.
`report_control`	Print a report of the control structure to stdout.
`report_info`	Print a report of the UMFInfo structure.
`report_numeric`	Print a report of the numeric factorization.
`report_symbolic`	Print a report of the symbolic factorization.
`slogdet`	Return the determinant of the matrix as (sign, logabsdet).
`solve`	Solve a linear system using the LU factorization.