并行单元归结

Parallel Unit Resolution

给出了基于神经网络的单元归结算法.首先将子句集S表示为δ形,并且用算子对(⊙, )引入两种类型的神经元,然后用这两种神经元构造子句集S的神经网络结构,而后给出基于Horn子句集的神经网络的归结算法,最后证明了该算法的完备性,并用实例进行了验证.

In this paper, a unit resolution based on neural network is proposed. We first express a clause set S in σ-form and introduce two kinds of neurons by a couple of operator (⊙,). Then we use the two neurons to construct a neural network structure of clause set S, and give the resolution algorithm on NN for Horn clause set S. Finally, we prove the completeness of the algorithm based on this resolution and give a practical example to check the theory.