Learning with risks based on M-location
Machine Learning
2023-12-01 v2 Machine Learning
Abstract
In this work, we study a new class of risks defined in terms of the location and deviation of the loss distribution, generalizing far beyond classical mean-variance risk functions. The class is easily implemented as a wrapper around any smooth loss, it admits finite-sample stationarity guarantees for stochastic gradient methods, it is straightforward to interpret and adjust, with close links to M-estimators of the loss location, and has a salient effect on the test loss distribution.
Cite
@article{arxiv.2012.02424,
title = {Learning with risks based on M-location},
author = {Matthew J. Holland},
journal= {arXiv preprint arXiv:2012.02424},
year = {2023}
}
Comments
Substantial update to initial version; refined theory, improved exposition, added experimental analysis