1 code implementation • 2 Sep 2021 • Rhys Green, Matthew Rowe, Alberto Polleri
This approach differs from standard calibration methods that use a global point prediction model as a starting point for the confidence estimate.
BIG-bench Machine Learning