Fixed Parameter Approximation Scheme for Min-max $k$-cut

TL;DR

This paper studies the NP-hard Minmax k-cut problem, introduces a parameterized approximation scheme based on the optimal value, and extends techniques to related partitioning problems.

Contribution

It provides the first parameterized approximation scheme for Minmax k-cut, with an exact algorithm running in fixed-parameter time, and generalizes to other norm-based partitioning measures.

Findings

NP-hardness and W[1]-hardness when parameterized by k.

A fixed-parameter approximation scheme with runtime depending on the optimal value.

Extension of techniques to minimize -norm measures for k-partitions.

Abstract

We consider the graph $k$-partitioning problem under the min-max objective, termed as Minmax $k$-cut. The input here is a graph $G=(V,E)$ with non-negative edge weights $w:E\rightarrow \mathbb{R}_+$ and an integer $k\geq 2$ and the goal is to partition the vertices into $k$ non-empty parts $V_1, \ldots, V_k$ so as to minimize $\max_{i=1}^k w(\delta(V_i))$. Although minimizing the sum objective $\sum_{i=1}^k w(\delta(V_i))$, termed as Minsum $k$-cut, has been studied extensively in the literature, very little is known about minimizing the max objective. We initiate the study of Minmax $k$-cut by showing that it is NP-hard and W[1]-hard when parameterized by $k$, and design a parameterized approximation scheme when parameterized by $k$. The main ingredient of our parameterized approximation scheme is an exact algorithm for Minmax $k$-cut that runs in time $(\lambda k)^{O(k^2)}n^{O(1)}$, where $\lambda$ is value of the optimum and $n$ is the number of vertices. Our algorithmic technique builds on the technique of Lokshtanov, Saurabh, and Surianarayanan (FOCS, 2020) who showed a similar result for Minsum $k$-cut. Our algorithmic techniques are more general and can be used to obtain parameterized approximation schemes for minimizing $\ell_p$-norm measures of $k$-partitioning for every $p\geq 1$.