【作者】 李永敏； 吉吟东； 朱善君； 孙增圻；

【Author】 Li Yongmin ’ Ji Yindong Zhu Shanjun (Department of Automation, Tsinghua University) Sun Zengqi (Department of Computer Science and Technology, Tsinghua University)

【机构】 清华大学自动化系； 清华大学计算机系；

【摘要】 本文提出了一种基于规则的前馈神经网络的设计方法和学习算法,它将粗糙集理论的定性分析能力和多层前馈神经网络的逼近能力结合在一起,得到一种结构可解释,计算简单、收敛速度快的前馈神经网络模型．在建立网络时,可以运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,然后根据这些规则可以确定网络中相应的隐层节点;网络的参数可以用BP算法迭代求出．更多还原

【Abstract】 The designing method and learning algorithm of a rule based Feedforward Neural Network is presented in this paper, which combine the powerful qualitative analysis capability of Rough Set theory and the outstanding approaching ability of multi-layer feedforward neural networks. The main advantages of this kind of neural network are its explainable structure, its simple computation and its exact accuracy. When constructing the network, one can find a set of rules from the given training data in terms of Rough Set theory, then create the neurons in the hidden layer according to those rules. The arguments of the neural network can be learnt with BP algorithm.更多还原