Fix scaling-dependence of termination criterion #47
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
|
Once conflicts are resolved I will re-review. |
The termination criterion for bisecting alpha implicitly assumes the optimal alpha to be in the order of magnitude of 1.0. For larger values of alpha, lower and upper can never get close enough, meaning the bisection never terminates successfully. For smaller values of alpha, the bisection may terminate much too early. This fixes that by checking the relative instead of the absolute difference using numpy.isclose. Also, this adds tests with a large and a tiny optimal alpha that would have failed with the previous termination criterion (they also wouldve failed for much more moderate scaling factors like 100).
mjacobse
force-pushed
the
fix_termination_criterion
branch
from
d8256ad to
33bf675
Compare
conflicts resolved |
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Outdated suggestions cannot be applied.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.
The termination criterion for bisecting alpha implicitly assumes the optimal alpha to be in the order of magnitude of
1.0. For larger values of alpha,loweranduppercan never get close enough, meaning the bisection never terminates successfully. This happens already with an example like this:For smaller values of alpha, the bisection may terminate much too early, returning a very suboptimal alpha.
This fixes that by checking the relative instead of the absolute difference using
numpy.isclose. Also, this adds tests with a large and a tiny optimal alpha that would have failed with the previous termination criterion.An alternative might be to check
np.nextafter(lower, upper) == upper, to ensure finding the exact threshold.See also #18 . A user-defined tolerance as suggested there might be a good future addition. But for now, this change would at least make the function usable for larger alphas.