Calibration estimators in ML.NET #1622
Description
In the discussions for #1579, one of the proposals was to introduce calibration estimators in ML.NET.
(from #1579)
.... a calibrator is just one peculiar form of trainer: it learns a monotonous function that transforms 'scores' into 'probabilities', with the goal to minimize the log-loss against the 'target label'. So, it is actually a univariate classification trainer. We should create a PlattCalibrationEstimator to train Platt calibrators and a PavCalibrationEstimator to train PAV calibrators.