摘要
在知识的广义定义的基础上,指出了数量性知识、符号性知识、样本性知识是主要的异构形态知识,分析了它们的不同特征.提出通过建立知识描述形式之间的转换联系达到知识集成的目的.分别给出了从神经网络到规则和从规则到框架的转换方法和具体步骤.通过一个机床加工选择的实例,验证了本方法的有效性.
Based on the generalized definition of knowledge, it is pointed out that quantity and symbol as well as sample knowledge are three main isomeric forms of knowledge. Their different characteristics are analyzed. An approach to knowledge integration by transformation connection among different forms of knowledge is proposed. The transformation methods and the implementation steps from neural networks to rules and from rules to frames are given respectively. The former conducts sample learning from original samples and generates the complete sample set to acquire the corresponding rules. The latter generates frames by recurrent way and simplifies the frame system obtained by utilizing inheritance relation. The effectiveness of the method is verified through a practical example. The main stages acquiring original samples and their normalizing them and neural network learning are given in the example.
出处
《华中理工大学学报》
CSCD
北大核心
1999年第2期4-6,共3页
Journal of Huazhong University of Science and Technology
基金
国家高技术研究发展计划资助
关键词
人工智能
知识集成
神经网络
产生式规则
artificial intelligence
knowledge integration
artificial neural network
rule
production frame
sample