scipy.sparse.

diags#

scipy.sparse.diags(diagonals,offsets=0,shape=None,format=None,dtype=<objectobject>)[source]#

Construct a sparse matrix from diagonals.

Warning

This function returns a sparse matrix – not a sparse array.You are encouraged to usediags_array to take advantageof the sparse array functionality.

Parameters:
diagonalssequence of array_like

Sequence of arrays containing the matrix diagonals,corresponding tooffsets.

offsetssequence of int or an int, optional
Diagonals to set (repeated offsets are not allowed):

  • k = 0 the main diagonal (default)

  • k > 0 the kth upper diagonal

  • k < 0 the kth lower diagonal

shapetuple of int, optional

Shape of the result. If omitted, a square matrix large enoughto contain the diagonals is returned.

format{“dia”, “csr”, “csc”, “lil”, …}, optional

Matrix format of the result. By default (format=None) anappropriate sparse matrix format is returned. This choice issubject to change.

dtypedtype, optional

Data type of the matrix. Ifdtype is None, the outputdata type is determined by the data type of the input diagonals.

Up until SciPy 1.19, the default behavior will be to return a matrixwith an inexact (floating point) data type. In particular, integerinput will be converted to double precision floating point. Thisbehavior is deprecated, and in SciPy 1.19, the default behaviorwill be changed to return a matrix with the same data type as theinput diagonals. To adopt this behavior before version 1.19, usedtype=None.

Returns:
new_matrixdia_matrix

dia_matrix holding the values indiagonals offset from the main diagonalas indicated inoffsets.

See also

spdiags

construct matrix from diagonals

diags_array

construct sparse array instead of sparse matrix

Notes

Repeated diagonal offsets are disallowed.

The result fromdiags is the sparse equivalent of:

np.diag(diagonals[0],offsets[0])+...+np.diag(diagonals[k],offsets[k])

diags differs fromdia_matrix in the way it handles off-diagonals.Specifically,dia_matrix assumes the data input includes padding(ignored values) at the start/end of the rows for positive/negativeoffset, whilediags assumes the input data has no padding.Each value in the inputdiagonals is used.

Added in version 0.11.

Examples

>>>fromscipy.sparseimportdiags>>>diagonals=[[1.0,2.0,3.0,4.0],[1.0,2.0,3.0],[1.0,2.0]]>>>diags(diagonals,[0,-1,2]).toarray()array([[1., 0., 1., 0.],       [1., 2., 0., 2.],       [0., 2., 3., 0.],       [0., 0., 3., 4.]])

Broadcasting of scalars is supported (but shape needs to bespecified):

>>>diags([1.0,-2.0,1.0],[-1,0,1],shape=(4,4)).toarray()array([[-2.,  1.,  0.,  0.],       [ 1., -2.,  1.,  0.],       [ 0.,  1., -2.,  1.],       [ 0.,  0.,  1., -2.]])

If only one diagonal is wanted (as innumpy.diag), the followingworks as well:

>>>diags([1.0,2.0,3.0],1).toarray()array([[ 0.,  1.,  0.,  0.],       [ 0.,  0.,  2.,  0.],       [ 0.,  0.,  0.,  3.],       [ 0.,  0.,  0.,  0.]])
On this page