Unlike heuristics, approximation algorithms have provable performance guarantees ... linear programming (LP), semidefinite programming (SDP), LP duality, randomized rounding, and primal-dual analysis.