一个基于符号神经网络的知识获取系统  

A Knowledge Acquisition System Based on Symbolic Neural Network

作　　者：王继成[1] 

机构地区：[1]同济大学计算机系,上海200092

出　　处：《电子学报》1998年第8期27-3,共1页Acta Electronica Sinica

基　　金：国家自然科学基金

摘　　要：本文提出了一种基于符号神经网络的知识获取方法，该方法首先用传统的机器学习方法获取关于某领域的粗略知识（coarseknowledge），然后把这些知识映射到神经网络结构，通过神经网络的自学习获取该领域的精细知识（fineknowedge）；并且采用该方法设计了一个基于符号神经网络的心电图分类的知识获取系统实验结果表明：该系统既解决了传统机器学习中知识精度、知识表示等问题；又解决了神经网络获取知识时间长。A method of knowledge aCquisition based on symbolic neural network is proud in which first the coarse knowledge of a field is acquired with the customary machine learning method, then the knowledge is mapped to neural network structure, at last the fine knowledge of a field is acquied through the auto-leaming of neural network. Also, the electrocardiogram classification knowledge acquisiton system based on symbolic neural network is designed with this method. Experiment results have shown that this system can solve the problems of knowledge refinement in machine learning method and the problems of long knowledge acquisition time and weak interpretation ability in neural network.

关 键 词：神经网络 机器学习 知识获取 符号系统 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]