摘要
在我们以前提出的短时记忆神经网络模型基础上［３］，我们在新模型中引入突触竞争机制，提出了一个新的短时记忆神经网络模型。模型仍由两个神经网络所组成；其一为与长时记忆共有的信息内容表达网络，另一个为指针神经元环路。由于表达区神经元与指针神经元间的突触权重的竞争，使得模型可以表现出由干扰引起的短时记忆的遗忘。相应于自由回忆序列位置效应和汉字组块两个心理学实验，对模型做了计算机仿真。仿真结果显示模型的行为与两个心理实验定量地符合得很好。由此表明现在的模型更合适于作为短时记忆的模型。
To the STM neural network model which we previously proposed[3], synaptic competition mechanism was introduced, and a new STM(short - term memory) neural network model was developed. This new model also consists of two neural networks; one is the network of representation of information content that share with long - term memory, the other is the pointer neuronal loop. Because of competition of synaptic weighs between neurons in representation area and neurons of pointer loop, the forgetting of STM caused by interference can be appeared in the model. Computer simulations of the model were performed in mimicking psychological experiments of the serial position effect of free recall and the chunking Chinese words. The results show that the behaviors of the model quantitatively fit well these psychological experiments. It verifies that the model described here is more suitable for modeling of STM.
出处
《生物物理学报》
CAS
CSCD
北大核心
1996年第4期603-608,共6页
Acta Biophysica Sinica
基金
国家基础研究"攀登计划"和国家自然科学基金
关键词
短时记忆
神经网络
神经网络模型
STM Neural network Synaptic competition Serial position curve Chunking of Chinese words
