The space of differences of convex functions on [0,1]
The author considers the space \(K[0,1]\) of differences of convex functions on the closed interval \([0,1]\) (they are continuous, maybe except at the endpoints). It turns out that \(K[0,1]\), under a specific norm \(\|\cdot \|\) (related to the biorthogonal functionals to the classical Schauder basis of \(C[0,1]\)) has a predual isometric to the space \(C(F)\) of continuous functions on the compact \(F=\{-1\} \cup [0,1] \cup \{2\}\). The constructed surjective isometry maps the positive cone of the space \(C(F)^*\) of regular Borel measures on \(F\) onto the cone of non-positive convex functions on \([0,1]\). A natural parallel is invoked by the classical fact that the dual of \(C[0,1]\) is the space \(BVN[0,1]\) of differences of increasing functions (with the total variation norm). The paper is concluded by providing an algorithm that, given some \(f\in K[0,1]\), constructs two non-positive convex functions \(g\) and \(h\) such that \(f=g-h\) and \(\|f\|=\|g\|+\|h\|\).
- scientific article; zbMATH DE number 3919065
- scientific article; zbMATH DE number 3903451
- A d.c.\ \(C^{1}\) function need not be difference of convex \(C^{1}\) functions.
- Some remarks on the space of differences of sublinear functions
- Representation of functions of several variables by difference of convex functions
This page was built for publication: The space of differences of convex functions on \([0,1]\)
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2752995)
