On semidefinite programming relaxations for graph coloring and vertex cover

  • Authors:
  • Moses Charikar

  • Affiliations:
  • Princeton University

  • Venue:
  • SODA '02 Proceedings of the thirteenth annual ACM-SIAM symposium on Discrete algorithms
  • Year:
  • 2002

Quantified Score

Hi-index 0.00

Visualization

Abstract

We investigate the power of a strengthened SDP relaxation for graph coloring whose value is equal to a variant of the Lovász ϑ-function. We show families of graphs where the value of the relaxation is 2 + ε for any fixed ε 0, yet the chromatic number is nδ for some fixed δ 0, which is a function of ε. This demonstrates the bound provided by the SDP is not strong enough to color a 3-colorable graph with no(1) colors.Kleinberg and Goemans considered an SDP relaxation for vertex cover whose value is n - ϑ1/2 (ϑ1/2 being the variant of the ϑ-function introduced by Schrijver). They asked whether this is within a ratio of 2 - ε of the optimal vertex cover for any ε 0. Our construction answers this question negatively.