The {\em Spanning Tree Congestion} problem is an easy-to-state NP-hard problem: given a graph $G$, construct a spanning tree $T$ of $G$ minimizing its maximum edge congestion where the congestion of an edge $e\in T$ is the number of edges $uv$ in $G$ such that the unique path between $u$ and $v$ in $T$ passes through $e$; the optimum value for a given graph $G$ is denoted $STC(G)$. It is known that {\em every} spanning tree is an $n/2$-approximation. A long-standing problem is to design a better approximation algorithm. Our contribution towards this goal is an $O(\Delta\cdot\log^{3/2}n)$-approximation algorithm for the minimum congestion spanning tree problem where $\Delta$ is the maximum degree in $G$. For graphs with maximum degree bounded by polylog of the number of vertices, this is an exponential improvement over the previous best approximation. For graphs with maximum degree bounded by $o(n/\log^{3/2}n)$, we get $o(n)$-approximation; this is the largest class of graphs that we know of, for which sublinear approximation is known for this problem. Our main tool for the algorithm is a new lower bound on the spanning tree congestion which is of independent interest. We prove that for every graph $G$, $STC(G)\geq \Omega(hb(G)/\Delta)$ where $hb(G)$ denotes the maximum bisection width over all subgraphs of $G$.
翻译:暂无翻译