视觉皮层突触连接网络演化模拟研究

发布时间：2018-05-12 10:04

本文选题：视觉皮层 + 复杂网络　；参考：《郑州大学》2016年硕士论文

【摘要】：大脑是目前最高效、最复杂、最稳定的系统之一,复杂网络已经成为研究大脑神经元连接和脑区连接的新手段。现有研究已经证明大脑网络具有小世界特性,其同时具备较短的平均路径长度与较高的聚类系数,分别对应大脑信息处理的功能分化与功能整合两个基本组织原则。视觉皮层存在于大脑的多个区域,是研究大脑的突破口。大脑内突触连接数量是按照一种先增后减的模式发展,大脑皮层的突触剪除机制可能是大脑网络高效稳健的关键所在,这也为创建高效率低能耗的网络提供新的参考方法。此外,神经系统中大多数的神经连接是短的,也存在少量长程连接。导致出现这种结构的内在机制还不清楚,同时考虑信息传递步数和总体布线代价的最小化,可能存在距离优先连接与大神经元优先连接的折衷。本文首先利用网络的数学描述验证视觉皮层网络的小世界特性,在此基础上研究视觉皮层的突触剪除机制与突触连接演化机制,研究内容概括如下:一、在大脑严格的空间限制下基于大脑体素坐标建立节点的位置信息,然后依据皮层组织结构特性,引入视觉皮层中节点间的空间距离的计算方式,建立节点间的连接概率函数的数学模型。在局部和全局范围内对视觉皮层网络进行模拟,验证了视觉皮层网络的小世界特性。二、以大脑皮层网络的突触剪除机制为启发,设计了网络减边演化算法。实验结果表明,该算法能够有效模拟突触剪除机制,并且演化后的网络会出现中心节点、小世界特性和高效率代价比等性质。验证了突触剪除机制可能是大脑网络高效性的关键所在这一推测。三、为模拟视觉皮层中总体布线代价和信息传递效率的权衡机制,结合复杂网络动态演化博弈理论,建立了基于距离与大神经元优先的网络演化博弈模型。实验结果显示,演化后的网络用长程边替代了一部分近邻边,改变了网络的拓扑结构和统计特征。演化博弈稳定于合作者占多数的状态时,网络出现相对较少的长程连接,这反应了距离优先机制。演化博弈稳定于完全背叛状态时,网络出现较多的长程连接,并会出现hub节点,这反应了大神经元优先机制。模型能够比较真实的模拟视觉皮层中布线代价和处理步骤最小化的折中关系。
[Abstract]:The brain is one of the most efficient, complex and stable systems at present. Complex networks have become a new method to study the connections of brain neurons and brain regions. It has been proved that the brain network has small world characteristics, which has shorter average path length and higher clustering coefficient, corresponding to the two basic organizing principles of functional differentiation and functional integration of brain information processing, respectively. The visual cortex, which exists in multiple regions of the brain, is a breakthrough in the study of the brain. The number of synaptic connections in the brain develops according to a pattern of first increasing and then decreasing. The mechanism of synaptic deletion in the cerebral cortex may be the key to the high efficiency and robustness of the brain network, which also provides a new reference method for the creation of high efficiency and low energy consumption networks. In addition, most neural connections in the nervous system are short and there are a few long-range connections. The internal mechanism leading to this kind of structure is not clear, and considering the minimization of the number of information transfer steps and the total routing cost, there may be a compromise between distance priority connection and large neuron priority connection. In this paper, we first use the mathematical description of the network to verify the small-world characteristics of the visual cortical network, and then study the mechanism of synaptic cutting and synaptic connection evolution of the visual cortex. The research contents are summarized as follows: 1. Under the strict space limit of the brain, the location information of the nodes is established based on the voxel coordinates of the brain, and then the spatial distance between the nodes in the visual cortex is calculated according to the structural characteristics of the cortex. The mathematical model of the connection probability function between nodes is established. The visual cortical network is simulated in the local and global scope to verify the small-world characteristics of the visual cortical network. Secondly, an evolutionary algorithm of network edge reduction is designed based on the synaptic cutting mechanism of cerebral cortex network. The experimental results show that the proposed algorithm can effectively simulate the mechanism of synapse cutting, and the evolved network will have some properties such as central nodes, small-world characteristics and high efficiency / cost ratio. It is proved that synaptic cutting mechanism may be the key to the efficiency of brain network. Thirdly, in order to simulate the tradeoff mechanism of overall routing cost and information transmission efficiency in visual cortex, a network evolutionary game model based on distance and large neuron priority is established based on the dynamic evolution game theory of complex network. The experimental results show that the evolved network uses long range edges to replace some of the nearest neighbor edges and changes the topological structure and statistical characteristics of the network. When the evolutionary game is stable in the state where the collaborators are the majority, there are relatively few long-distance connections in the network, which reflects the distance priority mechanism. When the evolutionary game is stable in the state of complete betrayal, there are more long-distance connections and hub nodes in the network, which reflects the mechanism of large neuron priority. The model can simulate the tradeoff between routing cost and processing steps in visual cortex.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：O157.5;R338

【相似文献】