Skip to content

Avoid using mean intercept for rmsle.#11588

Merged
trivialfis merged 4 commits into
dmlc:masterfrom
trivialfis:rmsle-intercept
Jul 26, 2025
Merged

Avoid using mean intercept for rmsle.#11588
trivialfis merged 4 commits into
dmlc:masterfrom
trivialfis:rmsle-intercept

Conversation

@trivialfis

@trivialfis trivialfis commented Jul 24, 2025

Copy link
Copy Markdown
Member

Close #11442 .

  • Split up the implementation.

@trivialfis
trivialfis requested a review from Copilot July 24, 2025 11:15

Copilot AI 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.

Pull Request Overview

This PR refactors the root mean squared logarithmic error (RMSLE) objective to avoid using mean intercept for regression. The changes address issue #11442 by creating a new implementation that inherits from FitIntercept instead of the shared RegLossObj class which uses FitInterceptGlmLike.

  • Extracts label validation logic into a standalone template function
  • Creates a new SquaredLogErrorRegression class that inherits from FitIntercept instead of using the shared RegLossObj
  • Removes unused includes and updates copyright years

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File Description
src/objective/regression_obj.cu Refactors RMSLE implementation with new class and extracted validation logic
src/objective/regression_loss.h Removes unused include and updates copyright
src/objective/quantile_obj.cu Removes unused includes and updates copyright
Comments suppressed due to low confidence (1)

src/objective/regression_obj.cu:119

  • The function call should be ValidateLabel<Loss> instead of ValidateLabel<Loss> since the extracted function is now a template function, but the template parameter syntax is inconsistent with the function definition which doesn't show template parameters.
      ValidateLabel<Loss>(this->ctx_, info);

Comment thread src/objective/regression_obj.cu Outdated
@trivialfis
trivialfis requested a review from hcho3 July 24, 2025 12:05
@trivialfis trivialfis mentioned this pull request Jul 24, 2025
10 tasks
@trivialfis
trivialfis merged commit 45da693 into dmlc:master Jul 26, 2025
61 of 62 checks passed
@trivialfis
trivialfis deleted the rmsle-intercept branch July 26, 2025 19:10
trivialfis added a commit to trivialfis/xgboost that referenced this pull request Jul 28, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Why did the default base_score for reg:squaredlogerror change in xgboost 3.0.0?

3 participants