Multi-Flip Networks: A Parallelization of local SAT{Algorithms

Local hill-climbing algorithms to solve the satis-ability problem have shown to be more eecient than complete systematic methods in many aspects. Many variants and reenements have been developed in the last few years. We now present a neural network approach to evaluate such local GSAT-like algorithms in a parallel manner, i.e. enlarging the neighbourhood of each possible move in the search space. We present an approach which allows the simultaneous change of truth value assignment for more than one variable at a time, such that the theoretical properties of the considered algorithms are preserved, and give experimental evidence that this algorithm is indeed faster than the respective sequential variants.