短时记忆的神经网络模型

SHORT TERM MEMORY NEURAL NETWORK MODEL

提出一个带有指针环路的短时记忆神经网络模型。模型包含两个神经网络，其中一个是与长时记忆共有的存贮内容表达网络，另一个为短时指针神经元环路。由于指针环路仅作为记忆内容的临时指针，因此，仅用很少的存贮单元即可完成各种短时记忆任务。计算机仿真证明：本模型确能表现出短时记忆的存贮容量有限和组块编码两个基本特征。

A neural network with pointer loop for short-term memory (STM) was proposed in this paper. The model consists of two neural networks; one is the network of representation of information content that share with long-term memory, the other is the STM pointer neuronal loop. Since the pointer loop is the temporal pointer of storage content, a few storage unitS are enough to perform various STM task. Computer simulations show that: the model have two fundamental characteristics of STM, limited capability and chunking.