Why is blindly transferring mindatainleaf tuning to a CatBoost symmetric-tree experiment invalid?
Classic MLLevel: IntermediateCatBoost · Training parameters
Why is blindly transferring min_data_in_leaf tuning to a CatBoost symmetric-tree experiment invalid?
Hints
- Check the policy-specific applicability.
- A familiar parameter name need not have universal scope.
Answer
The documented parameter applies to Lossguide and Depthwise growth, not SymmetricTree. Choose supported capacity controls for the active policy and record effective parameters rather than assuming every tree-library knob is interchangeable.
Source and licence: CatBoost · Training parameters · Apache-2.0 · answer: Quantoros editorial