A MinMaxScaler fitted on training data returns 1.3 for a new observation. Is it broken?
Classic MLLevel: FoundationExtrapolation after scaling
A MinMaxScaler fitted on training data returns 1.3 for a new observation. Is it broken?
Hints
- The fitted extrema are fixed.
- A linear map can extrapolate.
Answer
No. A value beyond the training maximum maps beyond the fitted range. Clipping is an explicit policy and loses information about the size of the excursion.
Source and licence: scikit-learn 1.7.2 — preprocessing · BSD-3-Clause · answer: Quantoros editorial