Why might sigmoid calibration be safer than a flexible nonparametric mapping on a small calibration set?
Classic MLLevel: IntermediateLow-data calibration
Why might sigmoid calibration be safer than a flexible nonparametric mapping on a small calibration set?
Hints
- Compare model flexibility with available calibration examples.
- Low variance comes with shape bias.
Answer
It estimates a low-dimensional sigmoid mapping, reducing variance but imposing shape assumptions. Validate those assumptions: simplicity helps sample efficiency, not universal correctness.
Source and licence: scikit-learn 1.7.2 — calibration · BSD-3-Clause · answer: Quantoros editorial