English

Max Cut and the Smallest Eigenvalue

Data Structures and Algorithms 2008-12-08 v5

Abstract

We describe a new approximation algorithm for Max Cut. Our algorithm runs in O~(n2)\tilde O(n^2) time, where nn is the number of vertices, and achieves an approximation ratio of .531.531. On instances in which an optimal solution cuts a 1ϵ1-\epsilon fraction of edges, our algorithm finds a solution that cuts a 14ϵ+8ϵo(1)1-4\sqrt{\epsilon} + 8\epsilon-o(1) fraction of edges. Our main result is a variant of spectral partitioning, which can be implemented in nearly linear time. Given a graph in which the Max Cut optimum is a 1ϵ1-\epsilon fraction of edges, our spectral partitioning algorithm finds a set SS of vertices and a bipartition L,R=SLL,R=S-L of SS such that at least a 1O(ϵ)1-O(\sqrt \epsilon) fraction of the edges incident on SS have one endpoint in LL and one endpoint in RR. (This can be seen as an analog of Cheeger's inequality for the smallest eigenvalue of the adjacency matrix of a graph.) Iterating this procedure yields the approximation results stated above. A different, more complicated, variant of spectral partitioning leads to an O~(n3)\tilde O(n^3) time algorithm that cuts 1/2+eΩ(1/\eps)1/2 + e^{-\Omega(1/\eps)} fraction of edges in graphs in which the optimum is 1/2+ϵ1/2 + \epsilon.

Keywords

Cite

@article{arxiv.0806.1978,
  title  = {Max Cut and the Smallest Eigenvalue},
  author = {Luca Trevisan},
  journal= {arXiv preprint arXiv:0806.1978},
  year   = {2008}
}
R2 v1 2026-06-21T10:49:47.670Z