14

Since Integer Linear Programming is NP-complete, there is a Karp reduction from any problem in NP to it. I thought this implied that there is always a polynomial-sized ILP formulation for any problem in NP.

But I've seen papers on specific NP problems where people write things like "this is the first poly-sized formulation" or "there is no known poly-sized formulation". That's why I'm puzzled.

Kaveh
  • 22,231
  • 4
  • 51
  • 111
andy
  • 253
  • 1
  • 5
  • 8
    You should point out an example or give a more complete quote ;) – hugomg May 15 '12 at 19:57
  • 1
    There exists a polynomial reduction from every NP-complete problem to every other NP-complete problem. However, just because we know one exists doesn't mean we know how to construct it. – Joe May 16 '12 at 05:05
  • @Joe we do know, all NP-completeness proofs I know are constructive. – andy May 16 '12 at 05:20
  • @andy You may have a constructive proof that 3-sat reduces to problem A, and that problem A reduces to problem B, and that 3-sat reduces to problem C, but a direct constructive proof that C reduces to B might not be known. – Joe May 16 '12 at 05:24
  • 3
    @Joe well, we know how to reduce any problem in NP to 3-sat, and also every practical NP-complete problem proof comes from a chain of reductions from 3-sat, so you can always compose reductions from any given NPC problem to any other. – andy May 16 '12 at 05:53
  • 10
    @andy didn't you just answer your question with that comment? You know every NP problem instance can be written as a polysized 3-SAT instance, and you know that a 3-SAT instance can be written as a polysized ILP instance, and polynomial applied to polynomial is another polynomial... what more do you expect from an answer? – Artem Kaznatcheev May 16 '12 at 15:43
  • @ArtemKaznatcheev perhaps the problem is that a ILP formulation of a problem should be "uniform" in some sense, or maybe I just misunderstood the authors (looking for an example for missingno I only found people using exponential formulations because they have good properties, but not because there aren't polynomial ones) – andy May 16 '12 at 16:07
  • @andy of course you could compose a chain of reductions like that. I guess I assumed you wanted something neater than reducing to 3-sat, otherwise why haven't you answered your own question? – Joe May 17 '12 at 20:54
  • I think the comments answer the question. @Joe why don't you post an answer? – Raphael May 18 '12 at 01:12
  • @andy: Could you give examples of such papers? Maybe they mean this is the first time a poly-size formulation has been used successfully or that there's no good poly-size formulation? – Opt May 18 '12 at 01:50
  • 2
    When someone says that this is the first poly-size formulation, what they mean is that it is the first explicitly given such formulation. The reductions obtained through SAT (even if one takes care of all details) do not look nice and are hard to work with. We usually want formulations that are natural and easy to work with. – Kaveh Jul 24 '12 at 09:05
  • Similarly, when one says that there is no polynomial size formulation, they probably mean that there is none that satisfies some other conditions. But to answer the question you should provide examples so we can look at the context and see what do authors exactly mean. – Kaveh Jul 24 '12 at 09:06

2 Answers2

5

This answer is mostly a recap of the comments on the question above.

If a problem is NP-complete, it can indeed be reduced to ILP, by using Karp's reductions (― Joe, andy). Claims of "polynomial sized formulations" from one problem to another, are likely meant as more direct formulations, as opposed to multiple reductions through SAT (― Kaveh).

Realz Slaw
  • 6,191
  • 32
  • 71
2

Yes. Every NP problem has a polynomial-sized ILP formulation.

Here is why. Every NP problem has a polynomial-sized formulation as an instance of SAT. Moreover, all of the usual boolean operators -- logical OR, logical AND, logical NOT, etc. -- can be expressed in ILP, using a constant number of variables and inequalities per boolean operator. See Express boolean logic operations in zero-one integer linear programming (ILP) for details of how to do that. Thus, we get at most a constant-sized blow-up when going from SAT to ILP. This implies that there is a polynomial-sized formulation of every NP problem as an ILP problem.

Community
  • 1
D.W.
  • 159,275
  • 20
  • 227
  • 470