English

Analysing Survey Propagation Guided Decimation on Random Formulas

Data Structures and Algorithms 2016-03-01 v1 Combinatorics

Abstract

Let Φ\varPhi be a uniformly distributed random kk-SAT formula with nn variables and mm clauses. For clauses/variables ratio m/nrk-SAT2kln2m/n \leq r_{k\text{-SAT}} \sim 2^k\ln2 the formula Φ\varPhi is satisfiable with high probability. However, no efficient algorithm is known to provably find a satisfying assignment beyond m/n2kln(k)/km/n \sim 2k \ln(k)/k with a non-vanishing probability. Non-rigorous statistical mechanics work on kk-CNF led to the development of a new efficient "message passing algorithm" called \emph{Survey Propagation Guided Decimation} [M\'ezard et al., Science 2002]. Experiments conducted for k=3,4,5k=3,4,5 suggest that the algorithm finds satisfying assignments close to rk-SATr_{k\text{-SAT}}. However, in the present paper we prove that the basic version of Survey Propagation Guided Decimation fails to solve random kk-SAT formulas efficiently already for m/n=2k(1+εk)ln(k)/km/n=2^k(1+\varepsilon_k)\ln(k)/k with limkεk=0\lim_{k\to\infty}\varepsilon_k= 0 almost a factor kk below rk-SATr_{k\text{-SAT}}.

Keywords

Cite

@article{arxiv.1602.08519,
  title  = {Analysing Survey Propagation Guided Decimation on Random Formulas},
  author = {Samuel Hetterich},
  journal= {arXiv preprint arXiv:1602.08519},
  year   = {2016}
}

Comments

arXiv admin note: substantial text overlap with arXiv:1007.1328 by other authors

R2 v1 2026-06-22T12:58:59.606Z