人工神经网络的发展及现状

The Development and Status of Artificial Neural Network

人脑由复杂的神经网络构成,拥有极强的信息处理、学习和记忆能力,并且具有功耗低、神经元密度高等特点。利用现有的电路元件搭建出类似大脑功能的人工神经网络成为当今人工智能研究的热点。在人工神经网络的研究中,信息的编码、神经元模型的建立以及突触的选择是最关键的3个要素。分别介绍了4种不同种类的编码方式和4种不同类型的神经元模型。对忆阻器用作神经突触的特点进行了简要阐述,并对其数学模型和对应的工作方式作了详细概括。

A human brain is a complicated neural network with low power consumption, high neuronal density and other unique characteristic, which is capable of complex information processing, learning and memorizing. Therefore, construction of artificial neural network by circuit component has become one of the most attractive topics of artificial intelligence. The information encoding, neuron modeling and synapsis selecting are the most crucial factors in the investigation of artificial neural network. Four different kinds of encoding rules and neuron models were introduced. Besides, the characteristics and advantages of the memristor that was served as synapsis were briefly illustrated. Its numerical models and the corresponding operation modes were discussed particularly.