[ML] Numerical robustness improvements for Bayesian Optimisation #2249
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
|
retest |
droberts195
reviewed
Apr 11, 2022
docs/CHANGELOG.asciidoc
Outdated
Comment on lines
61
to
62
| * Fix possible source of "x = NaN, distribution = class boost::math::normal_distribution<..." | ||
| log errors training classification and regression models. (See {ml-pull}2249[#2249].) |
There was a problem hiding this comment.
Since Easter is coming up and this still isn't merged it's likely to miss 8.2.0. If it does slip to a subsequent release then please remember to move the release note.
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Outdated suggestions cannot be applied.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.
This improves handling of numerical issues computing EI.
In particular, we weren't checking the condition of kernel covariance matrix. We don't need a lot of accuracy, but we need to handle edge cases, like singular matrices, NaNs, etc. In this case, zeroing the EI and its gradient is fine, the solver will fallback to taking random step.
Note that this is also a potential source of
[CTools.cc@152] x = NaN, distribution = class...log errors.