Item

Abstract

In this paper we consider the problem of computing minimum-cost spanning trees with depth restrictions. Specifically, we are given an $n$-node complete graph $G$, a metric cost-function $c$ on its edges, and an integer $k\geq 1$. The goal in the minimum-cost $k$-hop spanning tree problem is to compute a spanning tree $T$ in $G$ of minimum total cost such that the longest root-leaf-path in the tree has at most $k$ edges. Our main result is an algorithm that computes a tree of depth at most $k$ and total expected cost $O(\log n)$ times that of a minimum-cost $k$-hop spanning-tree. The result is based upon earlier work on metric space approximation due to Fakcharoenphol et al.[FRT03] and Bartal [Bart96,Bart98]. In particular we show that the problem can be solved exactly in polynomial time when the cost metric $c$ is induced by a so-called hierarchically well-separated tree.