sputils (Sparse Matrix Utilities)

Utility functions for scipy sparse matrices

rbf.sputils.add_rows(A, B, idx)

This adds the sparse matrix B to rows idx of the sparse matrix A

Parameters:

  • A ((n1, m) sparse matrix) – CSC, CSR, BSR, or COO matrix for best efficiency

  • B ((n2, m) sparse matrix) – CSC, CSR, BSR, or COO matrix for best efficiency

  • idx ((n2,) int array) – rows of A that B will be added to

Returns:

(n1, m) COO sparse matrix

rbf.sputils.divide_rows(A, x, inplace=False)

Divide the rows of the sparse matrix A by x. If inplace is True and A is CSC, CSR, BSR, or COO, then the operation is done in place.

Parameters:

  • A ((n, m) sparse matrix) – CSC, CSR, BSR, or COO matrix for best efficiency

  • x ((n,) float array) –

  • inplace (bool) – Whether to modify the data in A inplace. This is only possible if A is CSC, CSR, BSR, or COO.

Returns:

(n, m) COO sparse matrix

rbf.sputils.expand_cols(A, cols, cout, copy=False)

expand A along the columns.

Parameters:

  • A ((rin, cin) sparse matrix) – CSC, CSR, BSR, or COO matrix for best efficiency

  • cols ((rin,) int array) – The column in the output matrix that each column in A will be assigned to

  • cout (int) – The number of columns in the output matrix

  • copy (bool,) – Whether the data in the output matrix should be a copy of the data in A

Returns:

(rin, cout) COO sparse matrix

rbf.sputils.expand_rows(A, rows, rout, copy=False)

expand A along the rows.

Parameters:

  • A ((rin, cin) sparse matrix) – CSC, CSR, BSR, or COO matrix for best efficiency

  • rows ((rin,) int array) – The row in the output matrix that each row in A will be assigned to

  • rout (int) – The number of rows in the output matrix

  • copy (bool,) – Whether the data in the output matrix should be a copy of the data in A

Returns:

(rout, cin) COO sparse matrix

rbf.sputils.row_norms(A, order=2)

Computes the norm of each row in A. This is more memory efficient than scipy.sparse.linalg.norm because it does not make a full copy of the matrix

Parameters:

  • A ((n, m) float sparse matrix) – CSC, CSR, BSR, or COO matrix for best efficiency

  • order (int) –

Returns:

(n,) float array