M-convexity of the minimum-cost packings of arborescences

TL;DR

This paper demonstrates the discrete convexity properties of minimum-cost packings of arborescences and branchings, extending classical theorems and providing polynomial-time algorithms for related optimization problems.

Contribution

It proves the M-natural convexity of minimum-cost k-branchings and arborescences, extending Edmonds' theorem and linking to submodular base polyhedra, with applications to root location problems.

Findings

Minimum-cost k-branchings induce an M-natural convex function.

Minimum-cost k-arborescences are M-convex, forming a base polyhedron.

Polynomial-time solution for root location problem with M-natural convex cost functions.

Abstract

The aim of this paper is to reveal the discrete convexity of the minimum-cost packings of arborescences and branchings. We first prove that the minimum-cost packings of disjoint $k$ branchings (minimum-cost $k$-branchings) induce an $\mathrm{M}^\natural$-convex function defined on the integer vectors on the vertex set. The proof is based on a theorem on packing disjoint $k$-branchings, which extends Edmonds' disjoint branchings theorem and is of independent interest. We then show the $\mathrm{M}$-convexity of the minimum-cost $k$-arborescences, which provides a short proof for a theorem of Bern\'ath and Kir\'aly (SODA 2016) stating that the root vectors of the minimum-cost $k$-arborescences form a base polyhedron of a submodular function. Finally, building upon the $\mathrm{M}^\natural$-convexity of $k$-branchings, we present a new problem of minimum-cost root location of a $k$-branching, and show that it can be solved in polynomial time if the opening cost function is $\mathrm{M}^\natural$-convex.