基于SVM决策树的网络流量分类

提出一种用支持向量机(SVM)决策树来对网络流量进行分类的方法,利用SVM决策树在多类分类方面的优势,解决SVM在流量分类中存在的无法识别区域和训练时间较长的问题。对权威流量数据集进行了测试,实验结果表明,SVM决策树在流量分类中比普通的"一对一"和"一对多"SVM方法具有更短的训练时间和更好的分类性能,分类准确率可以达到98.8%。     

一种基于Dirichelt过程隐变量支撑向量机模型的目标识别方法

在目标识别中,对于样本数较多且分布复杂的数据,若将所有训练样本用来训练一个单一的分类器,会增加分类器的训练复杂度,且容易忽视样本的内在结构,不利于分类。因此人们提出了混合专家系统(ME),即将训练样本集划分为多个训练样本子集,并在每个子集上单独训练分类器。但是传统ME系统需要人为确定专家个数,并且每个子集的学习独立于后端的任务,如分类。该文提出一种基于Dirichlet过程(DP)混合隐变量(LV)支持向量机(SVM)模型(DPLVSVM)的目标识别算法,采用DP混合模型自动确定样本聚类个数,同时每个聚类中使用线性隐变量SVM(LVSVM)进行分类。不同于以往算法,DPLVSVM 将聚类过程和分类器的训练过程联合优化,保证了各个子集中样本的分布上的一致性和可分性,而且可以利用Gibbs采样技术对模型参数进行简便有效的估计。基于人工数据集、公共数据集以及雷达实测数据的实验验证了该文方法的有效性。     

机动目标声信号分类识别技术

针对机动目标声信号识别问题,对比三种分类方法的性能.在固定输入特征向量的前提下,将多分类问题细化为多个二分类问题,并分别训练二分类模型,通过投票的方式决策出目标类型.试验结果显示,支持向量机(Support Vector Machine,SVM)方法的准确率受样本数量影响较为明显,在样本数量充足的情况下,基于SVM构建...     

一种基于Hadoop平台CloudSVM的网络流量分类方法

大规模的netflow训练数据集是构建高质量、高稳定网络流量分类器的必然要求。但随着网络流特征维数的提高和数据集规模的扩大,无论是网络流的分析处理还是基于支持向量机(SVM)的分类器模型的训练,都无法在有效的时间内得到有效的处理结果。本文基于Hadoop云计算平台,采用MapReduce技术对SVM网络流量分类器进行分布式学习和训练,构建CloudSVM网络流量分类器。通过对来自校园网出口镜像的近2 T的大规模网络流量的跟踪文件的分布式存储和处理,对抽取的样本数据集进行分类,实验验证了基于Hadoop平台分布式存储和并行处理大规模网络数据集的高效率性,也验证了CloudSVM分类器在不降低分类准确度的情况下可以快速收敛到最佳,并随着大规模网络流样本的增加,SVM分类器训练的时间趋近平稳。     

卷积神经网络联合SVM的降水粒子分类

针对传统模糊逻辑降水粒子识别算法存在过度依赖专家经验来设置参数的缺陷,提出一种基于卷积神经网络（Convolutional Neural Network, CNN）和支持向量机（Support Vector Machine, SVM）联合结构的降水粒子分类方法。本文首先搭建了适用于双偏振雷达数据矩阵传播结构的4种卷积神经网络模型,通过对KOHX雷达各极化参数进行分块和堆叠操作,制作模型所需数据集并训练模型。然后根据各CNN模型对目标块的分类特点,结合SVM分类器,搭建能够识别5类目标粒子的联合结构。最后,对KOHX雷达0.5°仰角扫描数据进行测试,得到该联合结构模型的分类准确率达94.92%。并且对于不同仰角、不同雷达的扫描数据均能进行有效分类,表现出较好的鲁棒性。     

基于二叉树SVM的弹道目标RCS识别

二叉树支持向量机(SVM)是一种针对多类问题的有效分类器,具有结构简单、训练快的特点,但二叉树SVM容易出现误差积累,且不能输出识别结果的置信度。文中设计了一种基于隶属度计算的二叉树SVM分类器,首先,该分类器利用方差和最小准则选择节点,将多类问题转化为偏二叉树SVM分类问题,避免了误差积累,然后,利用特征变换空间的类中心和类半径,计算出样本结果的置信度,使得二叉树SVM分类器能够输出模糊结果。将上述二叉树SVM分类器应用于弹道目标的RCS特征识别,仿真结果表明了该方法的有效性。     

基于Matlab平台人脸识别仿真过程的探究

Matlab是一种科学与工程计算的高级语言,广泛地运用在包括信号与图像处理、控制系统设计等方面,我们使用Matlab作为平台,设计GUI,探究人脸识别的过程。在特征提取方面,运用主成分分析法(PCA)算法,对高维特征进行降维,保证了高位数据不失真,在分类器算法上采用支持向量机(SVM)和自适应提升(Adaboost)算法进行对比实验,SVM通过求解由全部训练样本对检测样本最佳线性表示的稀疏向量来进行分类,Adaboost算法针对不同的训练集训练同一个基本分类器(弱分类器)进行多次迭代,每次迭代增加错样本的权重,构成一个更强的最终的分类器(强分类器),实验结果表明,仿真能够达到较高的识别率和缩短识别的时间。     

基于迁移学习和SVM的糖网图像分类

为解决基于卷积神经网络(CNN)的糖尿病视网膜病变(DR)图像分类算法上普遍存在模型参数难以训练、易过拟合的问题,本文提出一种基于迁移学习和支持向量机(SVM)分类器的DR图像分类算法。首先,对DR图像进行预处理和数据扩增;其次,采用迁移学习方法预初始化深度学习分类算法中的经典框架VGGNet-16网络的模型参数,固定浅层网络参数不变,微调深层网络参数;最后,提取VGGNet-16最后一个隐藏层的特征向量训练支持向量机(SVM)分类器判定DR图像是否病变。实验结果表明,在Kaggle-DR公共数据集共35126张DR图像进行实验,在随机抽取的3500张作为测试集,分类准确率为0.931、敏感性为0.933、特异性为0.928,并能加快网络收敛和提高模型的泛化性。     

基于路径特征和支持向量机算法的硬件木马检测技术

硬件木马攻击成为当前集成电路(IC)面临的严重威胁。针对硬件木马电路具有隐蔽、不易触发以及数据集不均衡等特点，该文提出对门级网表进行静态分析的硬件木马检测技术。基于电路可测性原理建立涵盖节点扇入数、逻辑门距离、路径数、节点扇出数的硬件木马路径特征，简化特征分析流程；基于提取的路径特征，使用支持向量机(SVM)算法区分电路中的木马节点和正常节点。提出训练集双重加权技术，解决数据集不均衡问题，提升分类器的性能。实验结果表明，分类器可以用于电路中的可疑节点检测，准确率(ACC)达到99.85%；训练集静态加权有效提升分类器性能，准确率(ACC)提升5.58%；与现有文献相比，以36%的特征量，真阳性率(TPR)降低1.07%，真阴性率(TNR)提升2.74%，准确率(ACC)提升2.92%。该文验证了路径特征和SVM算法在硬件木马检测中的有效性，明确了数据集均衡性与检测性能的关系。     

一种基于SVM的雷达目标识别算法

针对目前支持向量机多类分类方法存在的缺点,丈中对支持向量机的高斯核函数进行改进,并提出一种结合留一法和单一验证法的参数选择新方案.基于3种雷达目标的HRRP数据,设计了相应的预处理算法,利用改进的SVM分类法用于高分辨距离像的雷达目标识别.从实验目标姿态稳定性、训练集大小稳定性和抗噪能力三个方面验证改进SVM的稳健性.     

GEMNET: a generalized, shuffle-exchange-based, regular, scalable,modular, multihop, WDM lightwave network

GEMNET is a generalization of shuffle-exchange networks and it can represent a family of network structures (including ShuffleNet and de Bruijn graph) for an arbitrary number of nodes. GEMNET employs a regular interconnection graph with highly desirable properties such as small nodal degree, simple routing, small diameter, and growth capability (viz. scalability). GEMNET can serve as a logical (virtual), packet-switched, multihop topology which can be employed for constructing the next generation of lightwave networks using wavelength-division multiplexing (WDM). Various properties of GEMNET are studied     

Thermodynamics of copper matte converting: Part III. Steady-state volatilization of Au,Ag, Pb,Zn, Ni,Se, Te,Bi, Sb,and As from slag,matte, and metallic copper

A comprehensive compilation of various thermodynamic data required for a complete analysis of copper matte converting reactions is presented. The data comprise estimated free energies of formation for such gases as SeO, SeS, TeO, TeS, BiO, BiS, SbO, SbS, AsO, and AsS, as well as activity coefficients in dilute copper alloys and vapor pressures of various elements and compounds. The volatilization of minor elements in steady-state reactors comprising gas and several condensed phases is mathematically formulated, and a parameter which governs the volatilization in such reactors is defined and named volatilization constant. The vapor pressures of various volatile species are calculated thermodynamically for the Noranda Process reactor by assuming equilibrium conditions. The volatilization constants of various minor elements are expressed explicitly as functions of oxygen and sulfur activities. Formerly Associate Professor, Department of Metallurgical Engineering, University of Utah, Salt Lake City, UT     

Bits, symbols, bauds, and bandwidth

The use of clear, concise, and unambiguous language in telecommunications engineering is vital to communicate a desired meaning and understanding. Terminology should be based on rigorous engineering principles and traceable to well known and dependable sources. This article deals with four everyday terms commonly encountered in the popular semi-technical press as well as in serious engineering periodicals. The misuse of these and other terms can lead to low-balled cost proposals, court litigation, and patent infringement cases. Imagine how an engineering student can be confused when she/he typically encounters bandwidth measured in bits per second     

Novel planar, square-shaped, dielectric-waveguide, single-, and dual-mode filters

This paper presents a novel concept of a square-shaped dielectric-waveguide resonator, which can easily realize a dielectric-waveguide cross-coupled filter or a dielectric-waveguide dual-mode filter, using the conventional printed circuit board (PCB) process. This planar dielectric-waveguide resonator has a higher Q value than a microstrip resonator. The physical mechanisms of both single- and dual-mode filters are elucidated. Some new coupling structures are developed for undergoing the PCB process. Filter design procedures are detailed. Design curves for the coupling coefficients of internal- and external-coupling structures are determined by full-wave finite-element-method electromagnetic calculations. Measurement results are highly consistent with theory for all trial filters. This study offers an effective means of producing low-cost high-performance planar circuit filters.     

PBX技术.市态.营销.选购

本文介绍了ＰＢＸ交换机的技术及市场需求态势。并提出ＰＢＸ产品的营销要素及其用户在选购时应遵循的原则。     

Dilemmas,uncertainty, risks,and board performance

This paper traces the origins of the words risk, uncertainty, governance and development and argues that a deeper understanding of each allows more effective decision taking in the boardroom. It argues that the role of the board of directors is to balance and rebalance continuously their irresolvable dilemma — ‘how do we drive our enterprise forward while keeping it under prudent control?’ It argues that it is the board’s role to focus on uncertainty, rather than risk, and this requires a different set of intellectual skills from board members to be able to cope with monitoring a range of diverse scenarios. This is crucial for a board to develop stronger ways of both leading their organisation and of ensuring the connectedness of the learning within and between the board and the operational unit’s risk taking. It advocates that to achieve this a board must develop new ways of learning — especially of thinking strategically and becoming more sensitised to the dynamics of their changing external environments. This will take them well beyond the comforts of their specialist managerial disciplines and into the true world of directing. The paper demonstrates the importance of continuous learning in reducing organisational uncertainty and risk. It uses the ‘Learning Board’ process and model as the central processor of both strategic and operational learning to create an annual rhythm for the board’s year.