【摘要】 提出了一种根据粗糙集理论进行ＢＰ网络设计的方法，它结合了粗糙集理论的强大的定性分析能力和ＢＰ网络的准确的逼近能力，得到一种可理解性好、计算简单、收敛速度快的神经网络模型．这种神经网络的学习算法的要点是：应用粗糙集的理论和方法，从给定学习样本数据中发现一组规则，并根据这些规则去建立网络模型中相应的隐层节点；然后用ＢＰ算法迭代求出网络的参数，从而完成网络的设计更多还原

【Abstract】 A designing method for BP neural networks based on rough set theory is presented in this paper. Rough set theory has a powerful capability for qualitative analysis, while BP neural networks can approach most problems with a much satisfying accuracy. By combining those advantages of the two theories, we can construct a kind of neural network with good understandability, simple computation and exact accuracy. The key idea of the learning algorithm for the proposed neural network is as follows: find a set of rules from the given training data by using Rough Set theory, construct the neurons in the hidden layers according to those rules, and then learn the arguments of the neural network with BP algorithm. The proposed neural network model has five layers: the first and the last layers represent input and output variables; the second layer stands for the discretization of input variables, and the fourth layer for discretization of output variables; each node in the third layer denotes a rule found by rough set theory, whose connections to the nodes of adjacent layers are determined by the conditions and conclusions of the rule. An example with satisfying results is also presented in this paper.更多还原