True if x is any CSR or COO sparse (array or matrix).

Returns a design matrix in CSR format. Note that only indices is passed, but not indptr because indptr is already precomputed in the call

4x3 CSR for matmul with 3x4.

Single nonzero at (row=0, col=col).

CSR matrix with one nonzero per row: row 0 → (col=0, val=1.0) row 1 → (col=_LARGE, val=2.0) ← argmax column for float comparison

Shared infra for make_splrep and make_splprep.

Matmul with at least one 1-D sparse operand. Runs on the ``(1, N)`` / ``(N, 1)`` backing with numpy 1-D matmul semantics: ``1d @ 1d``

Return the powers for each monomial in a polynomial. Parameters ---------- ndim : int Number of variables in the polynomial.

Split a 1-D index key into its real axis and new-axis counts. Returns ``(real_key, none_before, none_after)``: ``real_key`` is the si

Returns NVCC GPU code generation options.

Check we have complex conjugates pairs and reorder P according to YT, ie real_poles, complex_i, conjugate complex_i, .... The lexicograph

Returns a permuted range.

Differentiate polynomials represented with coefficients. p must be a 1-D or 2-D array. In the 2-D case, each column gives the coefficients o

10 ** x - 1 for x near 0

Raise square sparse ``A`` to ``power`` by exponentiation-by-squaring. Shared by :func:`matrix_power` and ``spmatrix.__pow__`` so the recursion

Return relative degree of transfer function from zeros and poles. This is simply len(p) - len(z), which must be nonnegative. A ValueErro

Pick the dtype for ``self.data * other`` / ``... / other``. numpy's natural promotion can land outside the cuSPARSE-supported set (e.

Sets up the deterministic random states of ``numpy`` and ``cupy``.

Collapse a ``(1, N)`` op result to 1-D. Only valid when the operation is genuinely 1-D (the 1-D-array ops run on the ``(1, N)`` backi