【摘要】 提出了一种基于神经网络的知识获取方法,该方法利用语言神经元,对具有开区域的连续输入变量,自动产生相应的语言变量输出,讨论了相应的神经网络训练和知识获取方法,所获取的知识以If-Then的规则形式表示,具有简洁、紧凑、不必进一步化简、易于理解等特点,并给出一个在教学型专家系统中获取专家领域知识的应用实例。更多还原

【Abstract】 A new methodology of extracting rules automatically from trained neur al networks is proposed.A linguistic neural unit,which has the ability to analyz e continuous input variables in open districts and produces corresponding lingu istic variables is described.The training method for network with linguistic ne ural layer is discussed.The rules extracted are represented in concise and comp rehensible If-Then forms ,and don’t need to be simplified further.An example showing how to extract rules automatically from the domain knowledge of educatio nal expert system is given.更多还原