`scipy.linalg._flapack:sgejsv`

Signature

def sgejsv(a, joba = 4, jobu = 0, jobv = 0, jobr = 1, jobt = 0, jobp = 1, lwork = Ellipsis, overwrite_a = 0)

Singular value decomposition by the Jacobi method, for high relative accuracy (LAPACK sgejsv).

LAPACK ships no complex counterpart, so only the real flavors exist.

a : ndarray: Matrix of shape (m, n) with m >= n.
joba : int, optional: Accuracy level, an index into 'CEFGAR': 0 through 5. Default is 4.
jobu : int, optional: Left singular vectors, an index into 'UFWN': 0 through 3. Default is 0.
jobv : int, optional: Right singular vectors, an index into 'VJWN': 0 through 3. jobv of 1 requires jobu below 2. Default is 0.
jobr : int, optional: If nonzero, extreme columns are pruned. Default is 1.
jobt : int, optional: If nonzero, the transpose may be factored instead. Default is 0.
jobp : int, optional: If nonzero, denormal values are flushed. Default is 1.
lwork : int, optional: Workspace size, at least 7. Default is the largest of the several bounds the routine documents.
overwrite_a : int, optional: If nonzero, a may be overwritten in place. Default is 0.

sva : ndarray: Singular values, length n.
u : ndarray: Left singular vectors; empty when jobu and jobt ask for none.
v : ndarray: Right singular vectors; empty when jobv and jobt ask for none.
workout : ndarray: The first seven workspace entries, which carry the scaling diagnostics.
iworkout : ndarray: The first three integer workspace entries: numerical rank, nullity and the count of accurate singular values.
info : int: 0 on success; if negative, the -info-th argument had an illegal value.