Glivenko-Cantelli for $f$-divergence
Statistics Theory
2025-03-25 v2 Machine Learning
Statistics Theory
Abstract
We extend the celebrated Glivenko-Cantelli theorem, sometimes called the fundamental theorem of statistics, from its standard setting of total variation distance to all -divergences. A key obstacle in this endeavor is to define -divergence on a subcollection of a -algebra that forms a -system but not a -subalgebra. This is a side contribution of our work. We will show that this notion of -divergence on the -system of rays preserves nearly all known properties of standard -divergence, yields a novel integral representation of the Kolmogorov-Smirnov distance, and has a Glivenko-Cantelli theorem. We will also discuss the prospects of a Vapnik-Chervonenkis theory for -divergence.
Keywords
Cite
@article{arxiv.2503.17355,
title = {Glivenko-Cantelli for $f$-divergence},
author = {Haoming Wang and Lek-Heng Lim},
journal= {arXiv preprint arXiv:2503.17355},
year = {2025}
}
Comments
26 pages, 1 figure