The N-R method of acquiring multi-step reasoning production rules based on NN

We present the N-R method of acquiring multi-step reasoning production rules based on a neural network (NN), so as to simulate one of the ways human beings learn-acquiring hierarchically structured knowledge from non-structured training samples. This method is composed of a characteristic NN structure to model the corresponding learning algorithm, and the transform algorithm which converts a trained NN to a reasoning network (RN), from which production rules can be obtained easily