Why does limiting boosting base trees to stumps restrict modeled interactions?
Classic MLLevel: IntermediateTree depth and interactions
Why does limiting boosting base trees to stumps restrict modeled interactions?
Hints
- Look at the variables used in one weak learner.
- Adding univariate functions does not create cross terms.
Answer
Each stump splits on one feature, so their additive sum represents additive single-feature effects. Deeper trees can place multiple features on a path and express interactions.
Source and licence: scikit-learn 1.7.2 — ensemble · BSD-3-Clause · answer: Quantoros editorial