Addendum to “ Approximation Algorithms for Unique Games ”

L. Trevisan

Abstract:We present a polynomial time algorithm that, given a unique game of value 1 − c/ log n, satisfies a constant fraction of constraints, where n is the number of variables. This improves an algorithm of Trevisan (ECCC TR05-34), that satisfies a constant fraction of constraints in unique games of value 1 − c/(log n) and, for sufficiently large alphabets, it improves an algorithm of Khot (STOC’02) that satisfies a constant fraction of constraints in unique games of value 1− c/(k(log k)), where k is the size of the alphabet. Our algorithm is based on semidefinite programming. The result presented in this note will be incorporated in a later version of ECCC TR05-34. 1 The Semidefinite Program A unique game [FL92, Kho02] is presented as a graph G = (V,E), a set S and a permutation πe : S → S for every edge e ∈ E. We think of a unique game as a constraint satisfaction problem where there is a variable for every vertex u, variables take values in the set S, and every edge e = (u, v) with associated permutation πe : S → S defines the constraint v = π(u,v)(u). The goal is to find an assignment that satisfies as many constraints as possible. The value of a unique game is the fraction of constraints satisfied by an optimal assignment. We consider the following integer programming formulation. We assume without loss of generality that the set S equals {1, . . . , |S|}. For every variable u of the unique game we have k := |S| boolean variables u1, . . . , uk in the integer program, with the intended meaning that if u = i then ui = 1 and uj = 0 for j 6= i. Each constraint v = f(u) contributes ∑ i∈S vf(i)ui to the objective function. The integer program, therefore, looks like this: max ∑