兰州大学机构库 >物理科学与技术学院
动态神经元模型的特性研究
Alternative TitleThe Study on the Characters of Dynamic Neuron Model
陈尚超
Thesis Advisor汪映海
2004-05-10
Degree Grantor兰州大学
Place of Conferral兰州
Degree Name硕士
Keyword动态神经元 模型 特性研究
Abstract神经系统作为一个极其复杂而高效的信息处理系统,其工作机制是众多自然科学基础学科研究的热点。相应地,这也促成了各种用于理论研究或实际应用的神经元模型以及神经网络模型的建立和发展。目前,神经元模型的研究正朝着两个不同的方向发展,一是保留其主要特性,用它作为基本的结构单元,构造实现一定功能的网络系统,即所谓的人工神经网络 [1];另一方面,就是在加入更详细的神经元生物学特性的基础上,提出更加符合生物学实际的神经元模型,即所谓的现实神经元模型[2]。 本篇论文重点讨论了近年来根据实际神经生物学实验结果所提出的一种神经元模型——动态神经元模型[3]。首先,引入该模型下的简单神经网络,重新分析了这种动态神经元网络的动力学性质及其对输入信号的响应情况,讨论了这种神经元模型的特点和有效性。 其次,我们对这种神经元模型在有噪声存在的情况下施加不同种类的输入信号,对各情况下的动力学特性进行了研究和分析。通过计算机模拟该模型对不同信号输入的响应,观察到外界噪声的存在对神经元的信息处理是有积极作用的,即在某些情况下,噪声的存在不但可以加强神经元对相对较弱信号的发现和检测,并且能够较为显著的提高动态神经元的发放能力。这也和实际的生物学实验结果相吻合。 然后,考虑到这种动态神经元模型在引入过程中的某些近似和省略,进一步把它和目前相对比较成熟,得到广泛研究的H-H神经元模型作了一些对比。我们主要分析了动态神经元与H-H神经元的发放频率和对直流刺激的关系,以及这两种神经元模型对于输入时序峰电位间隔信号的反馈情况。结果表明,前面讨论的动态神经元模型还存在着某些不足之处。 最后,对本文得到的结果以及可以进行的后续工作进行了总结和讨论。
Other AbstractIn most of the recent neuron models,they only regard synapse as the simple channel of temporal-spatial series and ignore the change of information between pre- and post-synapse. The output of neuron is usually S-type curve. S-type curve can show us the nonlinear and saturated properties. It is determined by steady firing rate at neurons,which implies only the steady property of neurons. However,real neurons are not always at a stationary firing state,and the output varies dynamically. Considering such properties, GU Fanji et al. suggested a new kind of neural network model with dynamic neurons [3]. In this thesis, we mainly focused on the study of this kind of dynamic neuron model based on recent experimental results. We use computer to simulate the dynamic neurons model and form a simple network. Then we input varies signals to study the dynamic and responding characters of the system. We then studied the above properties of this type of neural network model with the influence of external white noise. By using numerical simulations, we showed explicitly the effect of phase locking between the input and output of a single neuron. Moreover,our findings implies that the existence of external noise may enhance this type of neurons' ability to detect weak inputs signals and improve its firing efficacy. This result might have some useful meaning to explain the actual working mechanism of real neurons and might be good for the application in Artificial Intelligence. At last, we compare the Dynamic Neuron Model we discussed in this thesis with the famous H-H Neuron Model in two aspects: the feedback of input ISI and the responding firing rate to outside direct current. We find some shortcomings of the neuron model we have discussed in this thesis. This result can be beneficial for us to improve the current dynamic neuron model and furthermore, it’s helpful for us to find out some more realistic models.
URL查看原文
Language中文
Document Type学位论文
Identifierhttps://ir.lzu.edu.cn/handle/262010/230142
Collection物理科学与技术学院
Recommended Citation
GB/T 7714
陈尚超. 动态神经元模型的特性研究[D]. 兰州. 兰州大学,2004.
Files in This Item:
There are no files associated with this item.
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Altmetrics Score
Google Scholar
Similar articles in Google Scholar
[陈尚超]'s Articles
Baidu academic
Similar articles in Baidu academic
[陈尚超]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[陈尚超]'s Articles
Terms of Use
No data!
Social Bookmark/Share
No comment.
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.