StackedML
Practice
Labs
Questions
Models
Pricing
Sign in
Questions
/
Model Evaluation & Experimentation
/
Metrics (Classification)
/
Calibration
← Previous
Next →
407.
Isotonic Regression for Calibration
hard
Isotonic regression is an alternative to Platt scaling for calibration. What is its key advantage?
A
It is more data-efficient than Platt scaling and can be applied even with very small calibration sets
B
It directly optimizes the Brier score during calibration, producing better probabilistic forecasts than Platt scaling
C
It preserves the model's AUC-ROC during calibration while Platt scaling can reduce discriminative performance
D
It makes no parametric assumption about the shape of miscalibration and fits any monotonic mapping
Sign in to verify your answer
← Back to Questions