StackedML
Practice
Labs
Questions
Models
Pricing
Sign in
Questions
/
Model Evaluation & Experimentation
/
Metrics (Classification)
/
Calibration
← Previous
Next →
602.
Platt Scaling Mechanism
medium
Platt scaling is a post-hoc calibration method. What does it do?
A
It retrains the model's final layer using cross-entropy loss to directly optimize calibration
B
It rescales predicted probabilities linearly so they sum to 1 across all predicted classes
C
It applies isotonic regression to monotonically transform raw scores into calibrated probabilities
D
It fits a logistic regression on the model's raw scores to map them to calibrated probabilities
Sign in to verify your answer
← Back to Questions