PAC-Bayesian Bound for the Conditional Value at Risk. Mhammedi, Z.; Guedj, B.; and Williamson, R. C. In NeurIPS, 2020.
PAC-Bayesian Bound for the Conditional Value at Risk [link]Paper  PAC-Bayesian Bound for the Conditional Value at Risk [pdf]Pdf  abstract   bibtex   14 downloads  
Conditional Value at Risk (CVaR) is a family of "coherent risk measures" which generalize the traditional mathematical expectation. Widely used in mathematical finance, it is garnering increasing interest in machine learning, e.g., as an alternate approach to regularization, and as a means for ensuring fairness. This paper presents a generalization bound for learning algorithms that minimize the CVaR of the empirical loss. The bound is of PAC-Bayesian type and is guaranteed to be small when the empirical CVaR is small. We achieve this by reducing the problem of estimating CVaR to that of merely estimating an expectation. This then enables us, as a by-product, to obtain concentration inequalities for CVaR even when the random variable in question is unbounded.
@inproceedings{mhammedi2020pacbayesian,
title={{PAC-Bayesian Bound for the Conditional Value at Risk}},
author={Zakaria Mhammedi and Benjamin Guedj and Robert C. Williamson},
year={2020},
eprint={2006.14763},
booktitle = {NeurIPS},
abstract = {Conditional Value at Risk (CVaR) is a family of "coherent risk measures" which generalize the traditional mathematical expectation. Widely used in mathematical finance, it is garnering increasing interest in machine learning, e.g., as an alternate approach to regularization, and as a means for ensuring fairness. This paper presents a generalization bound for learning algorithms that minimize the CVaR of the empirical loss. The bound is of PAC-Bayesian type and is guaranteed to be small when the empirical CVaR is small. We achieve this by reducing the problem of estimating CVaR to that of merely estimating an expectation. This then enables us, as a by-product, to obtain concentration inequalities for CVaR even when the random variable in question is unbounded.},
archivePrefix={arXiv},
url = "https://arxiv.org/abs/2006.14763",
url_PDF = "https://arxiv.org/pdf/2006.14763.pdf",
primaryClass={cs.LG}
}
Downloads: 14