scipy.sparse.

block_diag#

scipy.sparse.block_diag(mats, format=None, dtype=None)[source]#

Build a block diagonal sparse matrix or array from provided matrices.

Warning

block_diag is switching to the sparse array interface.

For the case where no input arrays are sparse, this function is switching to returning a sparse array instead of sparse matrix. Control the sparse return class by making at least one input sparse, e.g., block_diag([coo_matrix(A), B]), or block_diag([coo_array(A), B]). That removes any deprecation warnings as well. For more general information about sparrays, see Migration from spmatrix to sparray. Handling of this no sparse input case will change no earlier than v1.20.

Parameters:

matssequence of matrices or arrays: Input matrices or arrays.
formatstr, optional: The sparse format of the result (e.g., “csr”). If not given, the result is returned in “coo” format.
dtypedtype specifier, optional: The data-type of the output. If not given, the dtype is determined from that of blocks.

Returns:

ressparse matrix or array: If at least one input is a sparse array, the output is a sparse array. Otherwise the output is a sparse matrix.

See also

block_array
diags_array

Notes

Added in version 0.11.0.

Examples

>>> from scipy.sparse import coo_array, block_diag
>>> A = coo_array([[1, 2], [3, 4]])
>>> B = coo_array([[5], [6]])
>>> C = coo_array([[7]])
>>> block_diag((A, B, C)).toarray()
array([[1, 2, 0, 0],
       [3, 4, 0, 0],
       [0, 0, 5, 0],
       [0, 0, 6, 0],
       [0, 0, 0, 7]])