神经元网络的共振效应—信息的检测与传导

发布时间：2018-03-21 05:20

本文选题：前馈神经元网络　切入点：皮层随机神经元网络　出处：《天津大学》2014年博士论文　论文类型：学位论文

【摘要】：大脑是人体最为复杂的系统，不同神经元和神经元网络是实现复杂脑功能的基础，信息在神经元及神经元网络中的检测和传导是神经科学中的一个重要问题。因此，本论文旨在通过构建外部刺激作用下的网络模型，包括前馈神经元网络、皮层随机神经元网络以及海马电阻网络，揭示外部刺激下，神经元网络神经编码的传导特性及共振等特性，进一步分析如何基于这些特性利用外部刺激对神经信息传导进行调控。本文首先构建了不同神经元模型组成的前馈神经元网络，分析网络中信息传导的基本规律。噪声在神经系统中是普遍存在的，其对神经系统中很多层面都会产生影响。研究发现，噪声对多层前馈神经元网络中的弱输入信号传导能力有增强作用，噪声诱导的随机共振成为弱信息的检测和传导的可能机制。从高频周期刺激信号模拟噪声环境所引发的振动共振研究中发现，前馈神经元网络中存在最优的高频刺激的幅值和频率使振动共振现象最为显著。通过分析网络参数对前馈神经元网络共振特性的影响，发现连接概率、突触时间常数、异质性等对微弱信息检测和传导产生一定的影响。其次，本文采用Izhikevich神经元模型构建了外部刺激作用下的皮层随机神经元网络模型，从振动共振的角度研究外部刺激作用下网络的同步和共振特性，以及突触可塑性对网络结构和功能的作用规律。研究发现，，高频刺激能够提高网络中弱信号的传导能力，网络的特性包括网络规模、突触连接概率以及突触权重等能够调节弱信号的传导。可塑性是与神经元网络学习相关的重要特性，通过对外部刺激下网络的放电规律进行分析发现，在没有外加交流刺激时，学习结束时网络表现出不相关的泊松放电行为。但是，当网络处于外加交流刺激时，学习过程结束时网络则表现出自激节律放电活动。此外，可塑性的存在使外部刺激的效果累积，即使是微小的改变也能积累产生明显的效果。最后，本论文建立了外部刺激作用下的海马CA3区椎体神经元两房室PR模型及相应的PR电阻耦合网络模型。研究发现，外部刺激参数不同时，单个神经元出现无放电、周期放电及无规则放电等放电模式。对网络同步特性研究结果表明，外部刺激能改变海马CA3区神经元网络的同步状态。本文基于同步和共振以突触可塑性研究了外部刺激下神经元网络信息的检测与传导，得到了高频刺激、噪声以及网络结构等对信息的检测与传导的影响规律。本文的研究成果为神经信息编码的调控方法及装置的研究提供了理论基础。
[Abstract]:The brain is the most complex system of human body. Different neurons and neural networks are the basis for the realization of complex brain functions. The detection and transmission of information in neurons and neural networks is an important issue in neuroscience. The purpose of this paper is to reveal the effects of external stimuli by constructing a network model, including feedforward neural network, cortical stochastic neural network and hippocampal resistor network. The conduction and resonance characteristics of neural network coding are analyzed. Based on these characteristics, how to regulate neural information transmission by external stimuli is further analyzed. In this paper, a feedforward neural network composed of different neuron models is constructed, and the basic rules of information transmission in the neural network are analyzed. Noise is ubiquitous in the nervous system. It has an effect on many levels of the nervous system. It has been found that noise enhances the ability of weak input signal transduction in multilayer feedforward neural networks. Noise induced stochastic resonance has become a possible mechanism for detecting and conducting weak information. The vibration resonance phenomenon is most obvious because of the optimal amplitude and frequency of high frequency stimuli in feedforward neural networks. By analyzing the influence of network parameters on the resonance characteristics of feedforward neural networks, the connection probability and synaptic time constant are found. Heterogeneity has a certain effect on weak information detection and transmission. Secondly, the cortical stochastic neural network model under external stimulation is constructed by using Izhikevich neuron model, and the synchronization and resonance characteristics of the network under external stimulus are studied from the view of vibration resonance. And the effect of synaptic plasticity on network structure and function. It has been found that high frequency stimuli can improve the transmission ability of weak signals in the network, and the characteristics of the network include the scale of the network. Synaptic connection probability and synaptic weight can regulate weak signal transduction. Plasticity is an important characteristic related to neural network learning. At the end of the learning process, the network exhibits an unrelated Poisson discharge behavior. However, when the network is stimulated by an additional AC, the network shows a spontaneous rhythmic discharge activity at the end of the learning process. The existence of plasticity accumulates the effects of external stimuli, even minor changes can accumulate obvious effects. Finally, a two-compartment PR model and a corresponding PR resistance-coupled network model of the spinal neurons in the CA3 region of hippocampus induced by external stimulation were established. It was found that there was no discharge in a single neuron at the same time when the external stimulation parameters were different. Periodic and irregular discharge patterns. The results show that external stimuli can change the synchronization state of hippocampal CA3 neural networks. In this paper, the detection and transmission of neural network information under external stimuli are studied by synaptic plasticity based on synchronization and resonance, and high frequency stimuli are obtained. The effects of noise and network structure on the detection and transmission of information are studied. The research results in this paper provide a theoretical basis for the study of the regulation and control methods and devices of neural information coding.
【学位授予单位】：天津大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：R318.04;TN911.23

【参考文献】