Skip to content

Fit the stacking meta-learner on out-of-fold base predictions - #66

Merged
ielbadisy merged 1 commit into
mainfrom
fix-stacking-oof
Nov 14, 2026
Merged

ielbadisy merged 1 commit into
mainfrom
fix-stacking-oof

Conversation

@ielbadisy

Copy link
Copy Markdown
Owner

The stacking learner fitted its meta-learner on in-sample base-learner predictions, which rewards overfitting base learners. Meta-features are now K-fold out-of-fold predictions (new spec entry folds, default 5, drawn from the current RNG so seed reproduces them); base learners are then refitted on all rows for prediction. The meta-features are kept at $state$meta_x. Documented in ?fit; tests added.

@ielbadisy
ielbadisy merged commit 1993bd6 into main Nov 14, 2026
@ielbadisy
ielbadisy deleted the fix-stacking-oof branch November 14, 2026 16:41
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.

1 participant