fips.linalg.SparseOperator

class fips.linalg.SparseOperator(matrix)

Bases: LinearOperator

A matrix held as a SciPy sparse matrix.

Parameters:

matrix (spmatrix | sparray) – The matrix. It is copied, as a CSR array (CSC for a CSC input), and the copy is read-only: changing matrix afterwards leaves the form as it was.

Raises:

TypeError – If matrix is not a SciPy sparse matrix.

Examples

A product of sparse forms stays sparse while at most a quarter of its entries are stored, and is dense past that:

>>> import scipy.sparse as sp
>>> from fips.linalg import SparseOperator
>>> A = SparseOperator(sp.diags([4.0, 3.0, 2.0, 2.0, 1.0, 1.0], format="csr"))
>>> A.diagonal()
array([4., 3., 2., 2., 1., 1.])
>>> type(A @ A).__name__
'SparseOperator'
>>> B = SparseOperator(sp.csr_matrix([[4.0, 1.0], [1.0, 3.0]]))
>>> type(B @ B).__name__
'DenseOperator'

Attributes

matrix

The matrix, as a SciPy sparse array whose arrays are read-only.

Methods

dense()	Return the form as a dense NumPy array.
diagonal()	Return the main diagonal, read from the stored entries.
logdet()	Return the log-determinant from a symmetric sparse LU factorization.
solve(B)	Solve self @ X = B by a sparse LU factorization (see LinearOperator.solve()).

Inherited

From LinearOperator: dot, sqrt, trace