一种改进的人工神经网络模型

提出一种新型人工神经网络模型，称为“基于模式神经元的人工神经网络（Pattern Neuron Based Artificial Neural Network，PNBANN）”。与现有的神经计算网络不同，PNBANN是一种完全基于神经元连接的网络模型。网络中的每一个神经元都唯一代表一种模式，每当接收新模式时，自动建立一个新的连接，把信息存储在网络中；而接收已有的模式时，已有的神经元连接得到加强。当模式神经元的输出达到所设定的感觉阈值时，对应模式的信息被记忆。因此，PNBANN就是不断地接收、存储各种信息，并把感觉足够强的模式记忆下来，这一过程更接近于人脑的学习、记忆过程。实验结果证明，PNBANN学习效率高，在学习新知识时不会影响已有的知识，同时具有很强的识别能力。

Modified artificial neural network model

In this paper,a new kind of artificial neural network is proposed,which is called Pattern-Neuron Based Artificial Neural Network(PNBANN).Different from the current neurocomputing networks,PNBANN is based on neuron connecting completely.Each neuron in PNBANN uniquely represents a pattern.Whenever an unknown pattern is received,a new neuron corresponding to this pattern is produced,and then the information is stored in the PNBANN.If a pattern has existed in the network when it is received,the connections included in the existing pattern neuron structure are enhanced.When the output of a pattern neuron exceeds the given feeling threshold,the pattern is memorized.Therefore,PNBANN can easily receive and store information constantly,and memorizes the information or patterns which can result strong feeling.This process is quite similar to the process of human brain to learn and keep something in mind.Simulations are made in this paper,and the results show that PNBANN can study with very high efficiency,and the existing patterns are not influenced while new pattern are incoming.It is also verified that PNBANN has a very high performance in recognition.