Optimal constant-time approximation algorithms and (unconditional) inapproximability results for every bounded-degree CSP

Yuichi Yoshida

Abstract: Raghavendra (STOC 2008) gave an elegant and surprising result: if Khot's Unique Games Conjecture (STOC 2002) is true, then for every constraint satisfaction problem (CSP), the best approximation ratio is attained by a certain simple semidefinite programming and a rounding scheme for it.In this paper, we show that similar results hold for constant-time approximation algorithms in the bounded-degree model. Specifically, we present the following: (i) For every CSP, we construct an oracle that serves an access, in constant time, to a nearly optimal solution to a basic LP relaxation of the CSP. (ii) Using the oracle, we give a constant-time rounding scheme that achieves an approximation ratio coincident with the integrality gap of the basic LP. (iii) Finally, we give a generic conversion from integrality gaps of basic LPs to hardness results. All of those results are unconditional. Therefore, for every bounded-degree CSP, we give the best constant-time approximation algorithm among all.A CSP instance is called ε-far from satisfiability if we must remove at least an ε-fraction of constraints to make it satisfiable. A CSP is called testable if there is a constant-time algorithm that distinguishes satisfiable instances from ε-far instances with probability at least 2/3. Using the results above, we also derive, under a technical assumption, an equivalent condition under which a CSP is testable in the bounded-degree model.