_{Crossposted on Mathematics SE}

I've seen in many optimisation papers the statement that general non-convex optimisation problem is NP-hard. If we assume that non-convex optimisation is in NP class, it can be shown, as I remember, that some NP-complete problems can be relaxed to corresponding continuous non-convex optimisation problems, which makes general non-convex optimisation problem NP-complete. But we assumed that it is in NP class, which is not obvious for me — indeed, it would mean that solution of multi-extremum problem can be verified in polynomial time. But it seems to be too strong and even practically important statement, which makes me sad that I do not know this fact (honestly, some important problems would be simpler that we think if it holds). In short — can one show that solution of non-convex optimisation problem can be verified in polynomial time, or it is in general more complicated problem than any in NP?