Should a small tabular CatBoost baseline always use Plain just because it trains faster?
Classic MLLevel: IntermediateCatBoost · Training parameters
Should a small tabular CatBoost baseline always use Plain just because it trains faster?
Hints
- Sample size changes the quality-speed trade-off.
- The fastest single run may not give the best validated model.
Answer
Not automatically. The documented Ordered scheme often improves small-data quality but may be slower. Compare validation quality and runtime under the same budget and check device/objective support before selecting the scheme.
Source and licence: CatBoost · Training parameters · Apache-2.0 · answer: Quantoros editorial