MAINT: optimize.bracket: don't fail silently #17704
Merged
Conversation
This file contains hidden or 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
mdhaber
commented
Jan 20, 2023
mdhaber
commented
Mar 4, 2023
This was referenced May 12, 2023
j-bowhay
reviewed
May 12, 2023
mdhaber
commented
May 12, 2023
j-bowhay
reviewed
May 12, 2023
scipy/optimize/_optimize.py
Outdated
| args : tuple, optional | ||
| Additional arguments (if present), passed to `func`. | ||
| grow_limit : float, optional | ||
| Maximum grow limit. Defaults to 110.0 | ||
| maxiter : int, optional | ||
| Maximum number of iterations to perform. Defaults to 1000. | ||
|
|
||
| Raises |
There was a problem hiding this comment.
Think this should be between returns and notes as per https://numpydoc.readthedocs.io/en/latest/format.html#sections
There was a problem hiding this comment.
Ugh I even looked at the standards to confirm; then I went and put it in the wrong place. I'll fix it.
scipy/optimize/_optimize.py
Outdated
| if np.any(np.isnan([xs, fs])): | ||
| x, fun = np.nan, np.nan | ||
| else: | ||
| imin = np.argmin(xs) |
There was a problem hiding this comment.
Possible mistake?
Suggested change
| imin = np.argmin(xs) | |
| imin = np.argmin(fs) |
There was a problem hiding this comment.
Unfortunately, I'm having trouble coming up with a realistic test case that exercises this with minimize_scalar. Sure, we could simulate calling _recover_from_backet_error with an artificial solve argument that raises a BracketError with out-of-order xs and fs and show that this picks the right one, but I don't think it's worth it.
j-bowhay
reviewed
May 13, 2023
j-bowhay
approved these changes
May 13, 2023
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.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Reference issue
Closes gh-14858
What does this implement/fix?
gh-14858 reported that
scipy.optimize.bracketmay silently return an invalid bracket, causing scalar minimization functions to fail. This PR resolves the issue by checking for failure before returning. It also begins to improve the function documentation and tests.Update:
bracketis used inbrent,golden, andminimize_scalarwith methods'brent'and'golden'. It's probably appropriate forbrentandgoldento raise errors whenbracketfails, butminimize_scalarshould terminate gracefully withsuccess=False. This requires plumbing the status information frombracketthrough the optimizers, which could this PR quite a bit more complex than intended - see a6aa40d. I ended up refactoring it that makes it much simpler, albeit a bit unorthodox.Additional information
In main,
bracketis not tested outside ofminimize_scalar, and the documentation is not explicit about the function's behavior. This is intended to be an improvement, but more work is certainly needed in followup PRs. For instance, this function still needs:minimize_scalar)For this PR, let's stick within the existing scope of fixing gh-14858 and these doc improvements so as not to delay progress.