A New Look at Survey Propagation and its Generalizations
This paper provides a new conceptual perspective on survey propagation, which is an iterative algorithm recently introduced by the statistical physics community that is very effective in solving random k-SAT problems even with densities close to the satisfiability threshold. We first describe how any SAT formula can be associated with a novel family of Markov random fields (MRFs), parameterized by a real number ρ\in [0,1]. We then show that applying belief propagation--a well-known ``message-passing'' technique for estimating marginal probabilities--to this family of MRFs recovers a known family of algorithms, ranging from pure survey propagation at one extreme (ρ= 1) to standard belief propagation on the uniform distribution over SAT assignments at the other extreme (ρ= 0). Configurations in these MRFs have a natural interpretation as partial satisfiability assignments, on which a partial order can be defined. We isolate cores as minimal elements in this partial ordering, which are also fixed points of survey propagation and the only assignments with positive probability in the MRF for ρ=1. Our experimental results for k=3 suggest that solutions of random formulas typically do not possess non-trivial cores. This makes it necessary to study the structure of the space of partial assignments for ρ<1 and investigate the role of assignments that are very close to being cores. To that end, we investigate the associated lattice structure, and prove a weight-preserving identity that shows how any MRF with ρ>0 can be viewed as a ``smoothed'' version of the uniform distribution over satisfying assignments (ρ=0). Finally, we isolate properties of Gibbs sampling and message-passing algorithms that are typical for an ensemble of k-SAT problems.