When XGBoost is trained as a correction to another model using basemargin, what must serving provide?
Classic MLLevel: IntermediateXGBoost · Prediction
When XGBoost is trained as a correction to another model using base_margin, what must serving provide?
Hints
- The learned trees correct an initial score.
- Reproduce the same initial-score mechanism.
Answer
The corresponding base margins for new rows under the same score/link convention, unless the full combination has been explicitly incorporated elsewhere. Dropping that offset changes the model being evaluated.
Source and licence: XGBoost · Prediction · Apache-2.0 · answer: Quantoros editorial