Nonuniform Graph Partitioning with Unrelated Weights

TL;DR

This paper presents improved approximation algorithms for the Minimum Nonuniform Partitioning problem, extending results to unrelated weights and dimensions, with better ratios for general and minor-excluded graphs.

Contribution

Introduces a bi-criteria approximation algorithm with improved ratios for nonuniform graph partitioning, including extensions to unrelated and multi-dimensional weights.

Findings

Achieves $O(\sqrt{\log n \log k})$ approximation for general graphs.

Attains $O(1)$ approximation for graphs with excluded minors.

Extends results to unrelated and multi-dimensional weight scenarios.

Abstract

We give a bi-criteria approximation algorithm for the Minimum Nonuniform Partitioning problem, recently introduced by Krauthgamer, Naor, Schwartz and Talwar (2014). In this problem, we are given a graph $G=(V,E)$ on $n$ vertices and $k$ numbers $\rho_1,\dots, \rho_k$. The goal is to partition the graph into $k$ disjoint sets $P_1,\dots, P_k$ satisfying $|P_i|\leq \rho_i n$ so as to minimize the number of edges cut by the partition. Our algorithm has an approximation ratio of $O(\sqrt{\log n \log k})$ for general graphs, and an $O(1)$ approximation for graphs with excluded minors. This is an improvement upon the $O(\log n)$ algorithm of Krauthgamer, Naor, Schwartz and Talwar (2014). Our approximation ratio matches the best known ratio for the Minimum (Uniform) $k$-Partitioning problem. We extend our results to the case of "unrelated weights" and to the case of "unrelated $d$-dimensional weights". In the former case, different vertices may have different weights and the weight of a vertex may depend on the set $P_i$ the vertex is assigned to. In the latter case, each vertex $u$ has a $d$-dimensional weight $r(u,i) = (r_1(u,i), \dots, r_d(u,i))$ if $u$ is assigned to $P_i$. Each set $P_i$ has a $d$-dimensional capacity $c(i) = (c_1(i),\dots, c_d(i))$. The goal is to find a partition such that $\sum_{u\in {P_i}} r(u,i) \leq c(i)$ coordinate-wise.