Max Cut is NP-complete #
Umbrella file for DescriptiveComplexity.MaxCut, Karp's MAX CUT
(Karp 1972) on the arc-marked vocabulary reused from
Feedback Arc Set, deriving NP-completeness from
DescriptiveComplexity.maxCut_sigmaSODefinable(DescriptiveComplexity.Problems.MaxCut.Membership): membership, by guessing one side of the cut and an injection of the marked pairs into the cut;DescriptiveComplexity.MaxCutRed.nae3Sat_ordered_fo_reduction_maxCut(DescriptiveComplexity.Problems.MaxCut.Reduction): hardness, from the NP-hardness of NAE-3SAT (DescriptiveComplexity.Problems.NaeThreeSat), and so ultimately from the Cook–Levin theorem, with no machine model anywhere.
No edge weights and no parallel edges are needed: the gadget graph is a plain
finite graph, its cut splits into three independently maximal families, and
the threshold is the Set.ncard of a marked binary relation. As with any
complexity-theoretic statement, these results are about finite structures only
(DescriptiveComplexity.ComplexityClass.mem_congr_finite/hard_congr_finite).
Max Cut is in NP: it is Σ₁-definable.
Dependency graph
Max Cut is NP-hard: NAE-3SAT, which is NP-hard, ordered-FO-reduces to it
by the gadget graph of DescriptiveComplexity.Problems.MaxCut.Interp.
Dependency graph
Max Cut is NP-complete, derived from the first-order reductions of this library and the Cook–Levin theorem.