BUG: sparse.linalg: mistake in unsymm. real shift-invert ARPACK eigenvalue selection #11199
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…value selection For real-valued unsymmetric problems, ARPACK can in some cases return one extra eigenvalue. The code that drops the extraneous eigenvalue did not take into account that in shift-invert mode, the which= keyword applies to the *transformed* eigenvalues, and dropped sometimes the wrong eigenvalue. Moreover, fix sort order in this case for 'LM', 'LR', 'LI' modes (largest items come first).
tylerjereddy
approved these changes
Dec 14, 2019
LGTM, test seems sensible and Azure coverage results look good for the affected code changes.
@denis-bz does this solve your problem?
We may want to merge soon-ish if this is "ok" since this is one of the two backports PRs under consideration for 1.4.0 "final," I think.
| assert_allclose(np.linalg.norm(A.dot(z) - z * w), 0, atol=np.sqrt(eps)) | ||
|
|
||
| # Check that the set of eigenvalues for `k` is a subset of that for `k+1` | ||
| dist = abs(prev_w[:,None] - w).min(axis=1) |
There was a problem hiding this comment.
This seems to be the check for the fix for the main complaint in #11198.
|
Ok this looks good to me & the original issue submitter, so I will merge & try to put this in 1.4.0 "final." Thanks Pauli! |
tylerjereddy
added a commit
to tylerjereddy/scipy
that referenced
this issue
Jan 20, 2020
* update SciPy 1.2.3 LTS release notes after backporting scipygh-10882, scipygh-11199, scipygh-10961, and scipygh-10833.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Reference issue
Closes: gh-11198
What does this implement/fix?
For real-valued unsymmetric problems, ARPACK can in some cases return
one extra eigenvalue. The code that drops the extraneous eigenvalue did not
take into account that in shift-invert mode, the which= keyword applies to the
transformed eigenvalues, and dropped sometimes the wrong eigenvalue.
Moreover, fix sort order in this case for 'LM', 'LR', 'LI' modes
(largest items come first), although arpack itself does not seem to guarantee
any order here.