QN solvers: Use different gradient norms for different for different loss functions. #4491
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
…ifferent loss functions
achirkin
added
2 - In Progress
achirkin
added
3 - Ready for Review
achirkin
changed the title
Codecov Report
@@ Coverage Diff @@
## branch-22.02 #4491 +/- ##
===============================================
Coverage ? 85.77%
===============================================
Files ? 236
Lines ? 19314
Branches ? 0
===============================================
Hits ? 16567
Misses ? 2747
Partials ? 0
Flags with carried forward coverage won't be shown. Click here to find out more. Continue to review full report at Codecov.
|
|
rerun tests |
1 similar comment
|
rerun tests |
cjnolet
approved these changes
Jan 28, 2022
|
rerun tests |
|
rerun tests |
1 similar comment
|
rerun tests |
tfeher
approved these changes
Feb 7, 2022
|
rerun tests |
|
@gpucibot merge |
vimarsh6739
pushed a commit
to vimarsh6739/cuml
that referenced
this pull request
Oct 9, 2023
…loss functions. (rapidsai#4491) Different loss functions may scale differently with the number of features. This has an effect on the convergence criteria. To account for that, I let a loss function define its preferred metric. As a result, the number of iterations should be less dependent on the number of features for all loss functions. Authors: - Artem M. Chirkin (https://github.com/achirkin) Approvers: - Corey J. Nolet (https://github.com/cjnolet) - Tamas Bela Feher (https://github.com/tfeher) URL: rapidsai#4491
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Outdated suggestions cannot be applied.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.
Different loss functions may scale differently with the number of features. This has an effect on the convergence criteria. To account for that, I let a loss function define its preferred metric. As a result, the number of iterations should be less dependent on the number of features for all loss functions.