【摘要】 针对传统专家系统知识获取的瓶颈性问题,提出了基于神经网络和传统知识获取与表示相结合的方法。传统的推理方式存在推理效率低和冲突消解问题,提出了由神经网络推理和逻辑推理所组成的混和推理系统。利用神经网络的自学习、自组织、自适应特点,来实现自动知识获取;混和推理既利用了神经网络的并行处理的效率、解决了传统推理存在的冲突消解问题,又克服了神经网络推理结果无法解释等特点,具有较高的准确性和效率性。最后给出了番茄病虫害诊断的应用实例。更多还原

【Abstract】 Knowledge acquisition (KA) is considered the bottleneck of the expert system-building process. To solve this problem,a combination of neural network with traditional knowledge expression and acquisition has been suggested and applied to the developed system. Traditional inference engine has some problems such as inference inefficient and conflict dispel. This paper advances a new method to solve these problems. The new inference engine consists of forward neural network inference and backward logical inference model. And at last the model of artificial neural network is applied in disease diagnosis expert system of tomato.更多还原