Skip to content

Add an experimental bouncing bounding method - #684

Merged
jrapin merged 1 commit into
masterfrom
bounce
May 20, 2020
Merged

Add an experimental bouncing bounding method#684
jrapin merged 1 commit into
masterfrom
bounce

Conversation

@jrapin

@jrapin jrapin commented May 16, 2020

Copy link
Copy Markdown
Contributor

Types of changes

  • Docs change / refactoring / dependency upgrade
  • Bug fix (non-breaking change which fixes an issue)
  • New feature (non-breaking change which adds functionality)
  • Breaking change (fix or feature that would cause existing functionality to change)

Motivation and Context / Related issue

Hard clipping on the bounds can lead to very detrimental loss of diversity. Bouncing on the border can help avoid this will being efficient and simple

How Has This Been Tested (if it applies)

Checklist

  • The documentation is up-to-date with the changes I made.
  • I have read the CONTRIBUTING document and completed the CLA (see CLA).
  • All tests passed, and additional code has been covered with new tests.

@jrapin jrapin self-assigned this May 16, 2020
@facebook-github-bot facebook-github-bot added the CLA Signed Do not delete this pull request or issue due to inactivity. label May 16, 2020
@jrapin
jrapin requested review from danthe3rd and teytaud May 16, 2020 08:06
@jrapin

jrapin commented May 18, 2020

Copy link
Copy Markdown
Contributor Author

@teytaud any thought on this? I find it's much better than clipping, it's closer to an actual constraint but way more efficient in high dimension

@jrapin

jrapin commented May 18, 2020

Copy link
Copy Markdown
Contributor Author

I would tend to even make this the default instead of clipping.
Related issue: #621

@danthe3rd danthe3rd left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks good to me from a code perspective :)

@jrapin

jrapin commented May 20, 2020

Copy link
Copy Markdown
Contributor Author

Merging since it does not induce any change for users

@jrapin
jrapin merged commit d212de5 into master May 20, 2020
@jrapin
jrapin deleted the bounce branch May 20, 2020 13:38
teytaud added a commit that referenced this pull request May 25, 2020
* firstMetaModel

* meta_model_for_shiva

* fix

* Type_fix:_.value_was_missing. Also better handling of exceptions.

* Tell not asked: I'm lost.

* Add recommendation.

* Add recommendation.

* Let us catch bad big values in metamodels.

* Deactivate BO test for windows

* Add metamodel in illcond tests.

* fixstaticanalysis

* Update optimizerlib.py

* updateMetaModel (#692)

* updateMetaModel

* Update optimizerlib.py

* Check suggest recommendation (#635)

* Add basic Pylint tests (#669)

* Allow repetitions for Choice parameters (#670)

* firstMetaModel

* meta_model_for_shiva

* Update post version number

* Fix repeated Choice hash (#676)

* fix

* Type_fix:_.value_was_missing. Also better handling of exceptions.

* Tell not asked: I'm lost.

* Add recommendation.

* Add recommendation.

* Let us catch bad big values in metamodels.

* Deactivate BO test for windows

* Add metamodel in illcond tests.

* fixstaticanalysis

* Lazy initialization in MLTuning (#673)

* Fix discrete mutation in dim 1 (#679)

Co-authored-by: Jeremy Rapin <jrapin@fb.com>
Co-authored-by: Teytaud <oteytaud@fb.com>

* fix

* Add an experimental bouncing bounding method (#684)

* Add a tutorial for windows users (#688)

* Update optimizerlib.py

* updateMetaModel (#692)

* updateMetaModel

* Update optimizerlib.py

Co-authored-by: Jeremy Rapin <jrapin@fb.com>
Co-authored-by: Jérémy Rapin <jrapin.github@gmail.com>
Co-authored-by: Fabien Teytaud <teytaud@lisic.univ-littoral.fr>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed Do not delete this pull request or issue due to inactivity. Difficulty: Low Priority: Medium Type: Enhancement New feature or request

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants