Modification of retinal ganglion cell axon morphology by prenatal infusion of tetrodotoxin

The cellular mechanisms by which the axons of individual neurons achieve their precise terminal branching patterns are poorly understood. In the visual system of adult cats, retinal ganglion cell axons from each eye form narrow cylindrical terminal arborizations restricted to alternate non-overlapping layers within the lateral geniculate nucleus (LGN). During prenatal development, axon arborizations from the two eyes are initially simple in shape and are intermixed with each other; they then gradually segregate to form complex adult-like arborizations in separate eye-specific layers by birth. Here we report that ganglion cell axons exposed to tetrodotoxin (TTX) to block neuronal activity during fetal life fail to form the normal pattern of terminal arborization. Individual TTX-treated axon arborizations are not stunted in their growth, but instead produce abnormally widespread terminal arborizations which extend across the equivalent of approximately two eye-specific layers. These observations suggest that during fetal development of the central nervous system, the formation of morphologically appropriate and correctly located axon terminal arborizations within targets is brought about by an activity-dependent process.     

基于Multi-Agent技术的三层信息融合系统研究

针对传统传感器网络管理复杂,系统信息融合智能化不高、精度低和模式单一、结构不清晰等不足,首先分析了Multi-Agent技术、传感器网络技术以及信息融合技术的独特优势,然后采用计算机网络分层结构思想和基于人工智能本体的知识表达理念,在信息融合过程中采用改进的SVM分类方法,构建了一种基于Multi-Agent技术的多传感器三层信息融合系统并对其具体融合过程进行了分析。最后对分类过程用MATLAB进行了分析。实验结果表明:系统分类精度较高,一定程度上不仅明显弥补了传统传感器的诸多不足,而且为后期决策提供了较为精准的目标参数。     

Embryonic assembly of a central pattern generator without sensory input

Locomotion depends on the integration of sensory information with the activity of central circuitry, which generates patterned discharges in motor nerves to appropriate muscles. Isolated central networks generate fictive locomotor rhythms (recorded in the absence of movement), indicating that the fundamental pattern of motor output depends on the intrinsic connectivity and electrical properties of these central circuits. Sensory inputs are required to modify the pattern of motor activity in response to the actual circumstances of real movement. A central issue for our understanding of how locomotor circuits are specified and assembled is the extent to which sensory inputs are required as such systems develop. Here we describe the effects of eliminating sensory function and structure on the development of the peristaltic motor pattern of Drosophila embryos and larvae. We infer that the circuitry for peristaltic crawling develops in the complete absence of sensory input; however, the integration of this circuitry into actual patterns of locomotion requires additional information from the sensory system. In the absence of sensory inputs, the polarity of movement is deranged, and backward peristaltic waves predominate at the expense of forward peristalsis.     

基于模糊划分的神经网络集成

基于模糊聚类思想,提出了一种神经网络集成方法.由训练数据的模糊聚类结果,把训练数据划分成相交子集,基于各子集生成集成的个体神经网络.由于各子集所包含的数据和数据的类别各不相同,因而个体神经网络性能和结构存在差异.子集个数确定集成中个体神经网络个数.另外,基于隶属度函数计算公式,提出了个体神经网络输出结论结合方法.理论分析和实验结果表明,此方法对模式分类能取得较好的效果.     

基于深度学习的MIMO系统联合信号检测与信道译码

为进一步提高多输入多输出(multiple-input multiple-output,MIMO)系统性能,研究了深度学习方法来联合解决MIMO系统信号检测与信道译码问题.通过将深度神经网络、自动编码器神经网络与传统MIMO通信系统的物理层架构进行有机融合,构建了基于神经网络的MIMO系统模型,可获取系统发射端的信息比...     

基于概率情景的多场点系统地震风险分析方法

该文针对传统地震危险性分析方法面向多场点系统(诸如某一区域内多个建筑、某个基础设施网络)整体地震风险计算时的局限性,提出了基于概率场景的多场点地震风险分析方法。该方法首先结合地震潜在震源的震级概率分布函数以及地震动预测公式,并考虑地震动空间相关性模型,通过随机模拟生成大量的概率地震情景(地震动分布图),在地震情景集及相应概率信息的基础上对一些典型的多场点系统,例如独立多场点系统、串联多场点系统、并联多场点系统等进行了系统风险分析。并通过对比分析单场点与多场点系统的地震风险,对比分析考虑和不考虑地震动估计误差空间相关性两种情况下的多场点系统地震风险,最终得出了三种类型多场点系统地震风险表现的不同特征。结果表明:不考虑空间相关性会导致独立多场点系统高估低水平损失的概率、低估高水平损失的概率;会导致串联系统风险估计整体偏高;会导致并联系统风险估计整体偏低。     

基于网络入侵检测的入侵者定位系统

提出并实现了一个基于网络入侵检测的入侵者定位系统．该系统利用分布在网络内的多个网络检测器检测网络攻击，并提取攻击指纹信息，由中央分析器对指纹信息进行相关性分析，构造攻击路径，从而得到入侵者的网络地址．和已有各种方法相比，该系统有更好的准确性、可扩展性和实时性，可有效地解决可控网络环境下的间接攻击问题．     

Neuropeptide fusion of two motor-pattern generator circuits

Animals make many different movements as circumstances dictate. These movements often involve the coordination of several neural networks, the output of which can be changed by modulatory substances. Here we report that the neuropeptide red pigment concentrating hormone modulates the interactions between two rhythmic pattern-generating networks in the lobster stomatogastric nervous system. Red pigment concentrating hormone markedly enhances the amplitude of synaptic interactions between elements of two pattern-generating networks--the cardiac sac and the gastric mill. Consequently, two networks operating under some circumstances virtually independently can be fused into one functional unit operating differently from either of the two original networks. These results show how a neuropeptide can alter the functional configuration of a neural network so that widely disparate outputs can be produced by the same neurons.     

Hierarchical structure and the prediction of missing links in networks

Networks have in recent years emerged as an invaluable tool for describing and quantifying complex systems in many branches of science. Recent studies suggest that networks often exhibit hierarchical organization, in which vertices divide into groups that further subdivide into groups of groups, and so forth over multiple scales. In many cases the groups are found to correspond to known functional units, such as ecological niches in food webs, modules in biochemical networks (protein interaction networks, metabolic networks or genetic regulatory networks) or communities in social networks. Here we present a general technique for inferring hierarchical structure from network data and show that the existence of hierarchy can simultaneously explain and quantitatively reproduce many commonly observed topological properties of networks, such as right-skewed degree distributions, high clustering coefficients and short path lengths. We further show that knowledge of hierarchical structure can be used to predict missing connections in partly known networks with high accuracy, and for more general network structures than competing techniques. Taken together, our results suggest that hierarchy is a central organizing principle of complex networks, capable of offering insight into many network phenomena.     

PKI技术及其在网络财务中的应用

本文就PKI（公开密钥基础设施）系统中的组成元素、技术以及功能分别阐述,从信息安全的基本知识出发,开发出一套最适合保证网络财务安全的方案和模式,并设计了一个基于PKI技术的网络财务系统,主要包括身份认证,SSL协议以及利用椭圆曲线加密体制的数字签名,解决了网络财务中的机密性、真实性、完整性和不可否认性等安全问题。     

智能大厦网络系统及其集成管理系统研究

认为智能大厦是运用系统集成控制中心通过综合布线系统将大楼内的 3A系统统一调控和管理 ,信息网络是智能化系统的集成控制中心 ,它具有各个智能化系统信息汇集和各类信息的综合管理功能 .智能大厦建设的本身是一个集成多种技术的复杂系统工程 ,其整个生命周期自用户需求分析开始 ,通过精心实施 ,直到评价、运行维护 ,应严格遵守系统工程的开发流程 .在综合分析智能大厦的概念、建造目的及基本组成的基础上 ,分析了智能大厦集成的综合网络系统结构 ,提出其拓扑形式以层次及网状结构为主 ,通常采用主干网络、子网结构进行设计 ,并重点以交换式快速以太网为例介绍了智能大厦主干网的构成 ,体现出快速以太网性能价格比高、升级容易、移植方便、易于扩展 ,是一种经济、实用并可向新技术过度的主干网方案 .最后对智能大厦集成管理系统进行了论述     

基于神经元连接的脑网络减边演化算法

神经元间的连接以先增大后减小的方式演化,这提示了大脑神经元网络在演化过程中存在“剪除”过程。以“剪除”为启发,本文首先使用数学方法对大脑网络进行建模,然后基于大脑网络中的神经元连接方式,设计了网络减边演化算法。最后,考虑布线消耗和信息传输之间的效率,建立了大神经元优先与距离优先的演化博弈模型,探索该模型对脑网络拓扑结构特性的影响。实验结果显示,在此演化过程中呈现了中心节点度≈200、小世界特性S>1和高效率代价比等性质, 表明该算法能够有效模拟仿神经元剪除机制。上述算法和模型为有效模拟高效低能耗的脑网络提供了一种新途径。     

基于深度学习的MIMO系统信道估计算法

精确的信道估计对于保证无线通信系统性能至关重要。针对多输入多输出(multiple input multiple output, MIMO)系统传统信道估计算法需已知信道统计信息以及性能与复杂度折中等问题,提出一种基于深度学习的多网络级联MIMO系统信道估计方案。基于卷积神经网络构建信道信息重建网络,初步重构出信道信息,进而基于深度残差网络构建信道估计网络进行级联得出估计结果,并利用多个损失函数对网络进行优化。仿真结果表明,在牺牲一定时间复杂度的情况下,所提方案的均方误差随信噪比增加逐渐优于线性最小均方误差(linear minimum mean squared error, LMMSE)估计算法,且不受信道统计信息的约束。     

新一代移动通信系统的发展前景

简要说明移动通信和个人通信迅速发展的趋势和欧洲第二代移动系统ＧＳＭ向新一人即第三代系统ＵＭＴＳ的变移和进化。简述ＵＭＴＳ的特点，包括陆地蜂窝网的进一步利用，业务与移动性层和交换与运输层的分开，及固定网与移动网的会聚。阐述当今无绳与蜂窝网移动电话的结合使用，有必要制成ＤＥＣＴ／ＧＳＭ双模手持机，新一代蜂窝风的多址方式急需从ＴＤＭＡ与ＣＤＭＡ作出选择，以及为了实现全球个人通信，必须考虑ＬＥＯ／ＭＥＯ卫     

基于信息网格的卫星网络仿真方法研究

针对目前卫星网络仿真系统中单一仿真软件很难同时满足卫星网络多方面的仿真需求,对比分析了卫星网络仿真软件的特点,提出了一种基于信息网格的卫星网络仿真方法,实现了一个开放的仿真框架。通过信息网格实现了不同卫星仿真软件之间的互连互通,通过实时共享仿真信息,增强了卫星网络仿真软件功能,为卫星网络仿真提供了一种新的解决方案。     

Dynamics of Drosophila embryonic patterning network perturbed in space and time using microfluidics

Biochemical networks are perturbed both by fluctuations in environmental conditions and genetic variation. These perturbations must be compensated for, especially when they occur during embryonic pattern formation. Complex chemical reaction networks displaying spatiotemporal dynamics have been controlled and understood by perturbing their environment in space and time. Here, we apply this approach using microfluidics to investigate the robust network in Drosophila melanogaster that compensates for variation in the Bicoid morphogen gradient. We show that the compensation system can counteract the effects of extremely unnatural environmental conditions--a temperature step--in which the anterior and posterior halves of the embryo are developing at different temperatures and thus at different rates. Embryonic patterning was normal under this condition, suggesting that a simple reciprocal gradient system is not the mechanism of compensation. Time-specific reversals of the temperature step narrowed down the critical period for compensation to between 65 and 100 min after onset of embryonic development. The microfluidic technology used here may prove useful to future studies, as it allows spatial and temporal regulation of embryonic development.     

控制系统的网络化发展

介绍了控制系统的网络化发展的三个阶段 ,预测了统一于TCP/IP协议的嵌入式设备网络控制系统的出现 .不同控制网络的集成化是当前控制系统网络化的主要特点之一 ,本文阐述了不同控制网络互连与集成的方法。从网络化发展的角度来看控制系统的发展 ,更有利于控制技术与计算机网络技术、通讯技术的融合 ,为实现控制和管理信息一体化系统提供新的思路。     

基于人工神经网络的旋转机械多故障同时性诊断策略

针对大型旋转机械多故障同时性诊断问题，基于人工神经网络，构造了一种由多个子网络组成的分级诊断网络（ＨＤＡＮＮ）。该网络旨在将一个大的分类模式空间划分为几个小的子空间，以便对各子网络进行有效的训练，提高各子网络的分类能力，从而使整个网络具有高精度的多故障同时性诊断能力，测试结果表明：ＨＤＡＮＮ网络不仅能准确地对单故障进行诊断，而且多故障同时存在的情况下，也能有效地识别出各种故障，该网络具有较高的诊断