Euclidean norm crashfix sparse matrices #40493
Conversation
src/sage/matrix/matrix2.pyx
Outdated
| A = self.dense_matrix().change_ring(CDF) | ||
| else: | ||
| A = self.change_ring(CDF) | ||
|
|
||
| A = A.conjugate().transpose() * A |
There was a problem hiding this comment.
while we're at it, maybe you can change this to .conjugate_transpose().
src/sage/matrix/matrix2.pyx
Outdated
| if self.is_sparse(): | ||
| A = self.dense_matrix().change_ring(CDF) | ||
| else: | ||
| A = self.change_ring(CDF) |
There was a problem hiding this comment.
overcomplicated, you could call .dense_matrix() unconditionally.
There was a problem hiding this comment.
True, the cost is null if the matrix is already dense.
|
actually there's #39831 that adds Either way, at least for sparse matrix over CDF, I think it's cleaner to implement although even with that, it may still be better for performance to call |
I see your point. I am unfamiliar with SVD and just wanted to patch the crash. |
Fixes #40492.
The call to
.norm()ornorm()with default parameter (p = 2for euclidean norm) crashes for sparse matrices.I go around it by checking if the matrix is sparse and if so, copy it into a dense matrix internally to compute the norm. This must be done since the euclidean norm uses the
.SVD()method which is not exposed for sparse matrices.📝 Checklist
⌛ Dependencies