English

Random 3CNF formulas elude the Lovasz theta function

Computational Complexity 2007-05-23 v1 Data Structures and Algorithms Logic in Computer Science

Abstract

Let ϕ\phi be a 3CNF formula with n variables and m clauses. A simple nonconstructive argument shows that when m is sufficiently large compared to n, most 3CNF formulas are not satisfiable. It is an open question whether there is an efficient refutation algorithm that for most such formulas proves that they are not satisfiable. A possible approach to refute a formula ϕ\phi is: first, translate it into a graph GϕG_{\phi} using a generic reduction from 3-SAT to max-IS, then bound the maximum independent set of GϕG_{\phi} using the Lovasz ϑ\vartheta function. If the ϑ\vartheta function returns a value <m< m, this is a certificate for the unsatisfiability of ϕ\phi. We show that for random formulas with m<n3/2o(1)m < n^{3/2 -o(1)} clauses, the above approach fails, i.e. the ϑ\vartheta function is likely to return a value of m.

Cite

@article{arxiv.cs/0603084,
  title  = {Random 3CNF formulas elude the Lovasz theta function},
  author = {Uriel Feige and Eran Ofek},
  journal= {arXiv preprint arXiv:cs/0603084},
  year   = {2007}
}

Comments

14 pages