Calibration and Uncertainty in Machine Learning: Advances, Limits, and Current Challenges
EN
Calibration and uncertainty quantification in machine learning have made great strides, yet they remain far from offering universal guarantees outside of stable environments. This analysis reviews their current state, the most common methods, and their main limitations.