中国组织工程研究

中国组织工程研究 ›› 2010, Vol. 14 ›› Issue (17): 3113-3116.doi: 10.3969/j.issn.1673-8225.2010.17.018

• 数字化骨科 digital orthopedics • 上一篇    下一篇

海马CA3区联想记忆功能的神经网络建模

赵旺兄，乔清理，王  丹   

  1. 天津医科大学生物医学工程系，天津市  300070
  • 出版日期:2010-04-23 发布日期:2010-04-23
  • 作者简介:赵旺兄★，女，1984年生，宁夏回族自治区隆德县人，汉族，天津医科大学在读硕士，主要从事神经信息学研究。 ssjy_2008@163.com

Neural network modeling of hippocampal CA3 associative memory functions

Zhao Wang-xiong, Qiao Qing-li, Wang Dan   

  1. Department of Biomedical Engineering, Tianjin Medical University, Tianjin  300070, China
  • Online:2010-04-23 Published:2010-04-23
  • About author:Zhao Wang-xiong★, Studying for master’s degree, Department of Biomedical Engineering, Tianjin Medical University, Tianjin 300070, China ssjy_2008@163.com

PDF

375

被引次数

5

摘要：

背景：海马是一个与记忆相关的重要脑区，在联想记忆中起着关键的作用。海马CA3区是形成联想记忆最重要的脑区，它从功能上被分为自联想记忆和异联想记忆。理解记忆的形成和读取需要建立海马功能的详细模型。
目的：根据海马CA3区的结构，建立一个海马CA3区功能的详细模型。
方法：该模型是一个三层的类Hopfield神经网络，由280个Izhikevich人工神经元组成。以Izhikevich神经元模型做为节点，用Hebbian学习法则调节权值。在给模型的输入加入白噪声的情况下用MATLAB对其仿真。在仿真中，记忆用同步放电序列表示。
结果与结论：仿真结果表明，模型的第三层具有异联想记忆功能；模型的第一层和第二层能完成自联想记忆功能。该模型很好的重现了海马CA3区三个子区域的记忆功能。但真实生物神经网络的功能极其动力学特性不可能通过建立一个简单的模型完全理解。该模型由280个神经元组成，远远小于实际神经元的数量，这也说明所建模型的特性与CA3实际生物神经网络的特性之间还存在很大差距。

关键词: 海马CA3功能, 神经网络, 自联想记忆, 异联想记忆, 神经元模型

Abstract:

BACKGROUND: Hippocampus is one of an important brain areas related with memory, and plays a critical role in associative memory function. Hippocampal CA3 is one of the most important regions to form associative memory. CA3 is functionally divided into autoassociative and heteroassociative memories, and memory formation and retrieval require the development of detailed models of hippocampal function.
OBJECTIVE: To establish a detailed model of hippocampal CA3 function according to CA3 structure.
METHODS: The model was a three-layered Hopfield-like neural network and was constituted by 280 Izhikevich artificial neurons, and is modulated by Hebbian rules. The model was simulated using MATLAB under the condition of adding the Gaussian white noise to its input. In the simulation, memories were represented by synchronous firing sequences.
RESULTS AND CONCLUSION: The simulating results show that the third layer of model had heteroassociative memory function; the first and the second layer of the model could implement autoassociative memory. The model implemented well the memory functions of three subregions of hippocampal CA3. But it is impossible to understand the functions and dynamics of a real biological neural network by constructing a simple model. The model proposed has 280 neurons, which are far less than the real number of neurons. It suggests that there is a big gap between the properties of the model and a real biological neural network of CA3.

中图分类号: