Can bin-wise scaling improve consistency and adaptivity of prediction uncertainty for machine learning regression ?
Abstract
Binwise Variance Scaling (BVS) has recently been proposed as a post hoc recalibration method for prediction uncertainties of machine learning regression problems that is able of more efficient corrections than uniform variance (or temperature) scaling. The original version of BVS uses uncertainty-based binning, which is aimed to improve calibration conditionally on uncertainty, i.e. consistency. I explore here several adaptations of BVS, in particular with alternative loss functions and a binning scheme based on an input-feature (X) in order to improve adaptivity, i.e. calibration conditional on X. The performances of BVS and its proposed variants are tested on a benchmark dataset for the prediction of atomization energies and compared to the results of isotonic regression.
Keywords
Cite
@article{arxiv.2310.11978,
title = {Can bin-wise scaling improve consistency and adaptivity of prediction uncertainty for machine learning regression ?},
author = {Pascal Pernot},
journal= {arXiv preprint arXiv:2310.11978},
year = {2023}
}
Comments
This version corrects an error in the estimation of the Sx scores for the test set, affecting Fig. 2 and Tables I-III of the initial version. The main points of the discussion and the conclusions are unchanged