A note on the richness of convex hulls of VC classes

Abstract

We prove the existence of a class $A$ of subsets of $\mathbb{R}^d$ of VC dimension 1 such that the symmetric convex hull $F$ of the class of characteristic functions of sets in $A$ is rich in the following sense. For any absolutely continuous probability measure $\mu$ on $\mathbb{R}^d$, measurable set $B$ and $\varepsilon \gt 0$, there exists a function $f$ in $F$ such that the measure of the symmetric difference of $B$ and the set where $f$ is positive is less than $\varepsilon$. The question was motivated by the investigation of the theoretical properties of certain algorithms in machine learning.