Towards efficient sampling: exploiting random walk strategies

From a computational perspective, there is a close connec-tion between various probabilistic reasoning tasks and the problem of counting or sampling satisfying assignments of a propositional theory. We consider the question of whether state-of-the-art satisfiability procedures, based on random walk strategies, can be used to sample uniformly or near-uniformly from the space of satisfying assignments. We first show that random walk SAT procedures often do reach the full set of solutions of complex logical theories. Moreover, by interleaving random walk steps with Metropolis transitions, we also show how the sampling becomes near-uniform.