Temperature Scaling vs Platt Scaling for Neural Classifiers
Temperature scaling needs one parameter; Platt scaling needs two.
Priya Subramaniam
Senior Editor & Staff Writer
Priya Subramaniam
Priya covers the theoretical foundations of calibration and probabilistic reasoning, drawing on a decade of work spanning academic machine learning research and applied AI consulting. She previously contributed to several peer-reviewed venues on uncertainty quantification before moving into science journalism.
2 stories
Bayesian vs Frequentist Uncertainty in Sequential Decision Making
Bayesian updating handles sequential uncertainty better than frequentist methods.
Priya Subramaniam