Nonlinear Formations and Improved Randomized Approximation Algorithms for Multiway and Multicut Problems

Abstract

We introduce nonlinear formulations of the multiway cut and multicut problems. By simple linearizations of these formulations we derive several well known formulations and valid inequalities as well as several new ones. Through these formulations we establish a connection between the multiway cut and the maximum weighted independent set problem that leads to the study of the tightness of several LP formulations for the multiway cut problem through the theory of perfect graphs. We also introduce a new randomized rounding argument to study the worst case bound of these formulations, obtaining a new bound of 2a(H)(1 - ) for the multicut problem, where ac(H) is the size of a maximum independent set in the demand graph H.