Unsatisfiability Bounds for Random CSPs from an Energetic Interpolation Method
The interpolation method, originally developed in statistical physics, transforms distributions of random CSPs to distributions of much simpler problems while bounding the change in a number of associated statistical quantities along the transformation path. After a number of further mathematical developments, it is now known that, in principle, the method can yield rigorous unsatisfiability bounds if one “plugs in an appropriate functional distribution”. A drawback of the method is that identifying appropriate distributions and plugging them in leads to major analytical challenges as the distributions required are, in fact, infinite dimensional objects. We develop a variant of the interpolation method for random CSPs on arbitrary sparse degree distributions which trades accuracy for tractability. In particular, our bounds only require the solution of a 1-dimensional optimization problem (which typically turns out to be very easy) and as such can be used to compute explicit rigorous unsatisfiability bounds.
KeywordsEnergy Function Interpolation Method Univariate Factor Degree Sequence Optimal Assignment
Unable to display preview. Download preview PDF.
- 1.Achioptas, D., Sorkin, G.: Optimal myopic algorithms for random 3-sat. In: Proceedings of the 41st Annual Symposium on Foundations of Computer Science 2000, pp. 590–600. IEEE (2000)Google Scholar
- 4.Achlioptas, D., Coja-Oghlan, A.: Algorithmic barriers from phase transitions. In: 49th Annual IEEE Symp. on Foundations of Computer Science 2008, pp. 793–802 (2008)Google Scholar
- 5.Achlioptas, D., Menchaca-Mendez, R.: Exponential lower bounds for dpll algorithms on satisfiable random 3-cnf formulas (2012) (to appear in SAT 2012)Google Scholar
- 6.Bayati, M., Gamarnik, D., Tetali, P.: Combinatorial approach to the interpolation method and scaling limits in sparse random graphs. In: STOC 2010, pp. 105–114 (2010)Google Scholar
- 8.Coja-Oghlan, A.: On belief propagation guided decimation for random k-sat. In: Proceedings of the Twenty-Second Annual ACM-SIAM Symposium on Discrete Algorithms, pp. 957–966. SIAM (2011)Google Scholar
- 17.Monasson, R., Zecchina, R.: Entropy of the K -satisfiability problem. Phys. Rev. Lett. 76, 3881–3885 (1996), http://link.aps.org/doi/10.1103/PhysRevLett.76.3881 MathSciNetMATHCrossRefGoogle Scholar