On the integrality gap of the maximum-cut semidefinite programming relaxation in fixed dimension

Abstract

We describe a factor-revealing convex optimization problem for the integrality gap of the maximum-cut semidefinite programming relaxation: for each $n\ge 2$ we present a convex optimization problem whose optimal value is the largest possible ratio between the value of an optimal rank-$n$ solution to the relaxation and the value of an optimal cut. This problem is then used to compute lower bounds for the integrality gap.

17 pages, 2 figures