Reoptimization of constraint satisfaction problems with approximation resistant predicates

Abstract

If k O n = ( ) log and a predicate P is approximation resistant for the reoptimization of the Max-EkCSP-P problem, then, after inserting a truth-value into the predicate and imposing some constraint, there exists a polynomial algorithm with the approximation ratio q P d Pis a "random" threshold approximation ratio of the predicate P. The ratio q P ( )is a threshold approximation ratio.