一种改进的二叉树多分支持向量机算法

二叉树支持向量机分类算法主要是构造一个偏二叉树或是构造一颗完全二叉树,但是偏二叉树分类的准确性虽高而分类的效率低,完全二叉树分类的效率高但是准确性不高。本文提出一种算法,结合了以上两种二叉树构造方法的优点,并且更能反映样本的真实分布。实验结果表明,新算法具有较高的推广性能。     

An improved curvilinear finite difference (CFD) method for arbitrary mesh systems

This paper describes the automatic computer generation of finite difference approximations for partial derivatives in arbitrary mesh systems using an “Improved CFD” method. In the first part of the paper, the algorithm of the proposed Improved CFD method, used to solve a general second-order linear partial differential equation defined in a two-dimensional domain, will be presented. The performance of this numerical method will then be compared with the “Original CFD” method developed by Lau[1], and a so-called least square surface fit method developed by Liszka and Orkisz[2]. In the proposed method, extensive use is made of matrix algebra. The method is therefore highly systematic and can be easily implemented into computer programmes. Numerical examples tested in this paper indicate that, for irregular meshes, better numerical accuracy can be attained by the proposed method as compared with the Original CFD method. At the same time, the straightforward extension of the present method to the generation of higher-order-than-two finite difference approximations and to the solution of three-dimensional field problems will also be demonstrated.In the second part of the paper, the computational procedures for the numerical solution of a first boundaryvalue problem which is governed by a single second-order nonlinear partial differential equation will be derived.     

Wspruce:一种改进的可用带宽测量方法

可用带宽是反映网络状态的主要参数,对其准确的估计与测量是流量工程和网络监测等必须解决的问题,但对它的实际测量存在许多困难。针对Spruce可用带宽测量方法存在收敛慢、开销大的问题,提出了一种Spruce可用带宽测量的改进方法（Wspruce）。该方法利用隐马尔可夫模型（HMM）的序列预测特性,可以对可用带宽做出较为准确的分析。实际测量结果表明,该方法在可用带宽测量中估算速度更快,开销更低。     

Cost curves: An improved method for visualizing classifier performance

This paper introduces cost curves, a graphical technique for visualizing the performance (error rate or expected cost) of 2-class classifiers over the full range of possible class distributions and misclassification costs. Cost curves are shown to be superior to ROC curves for visualizing classifier performance for most purposes. This is because they visually support several crucial types of performance assessment that cannot be done easily with ROC curves, such as showing confidence intervals on a classifier's performance, and visualizing the statistical significance of the difference in performance of two classifiers. A software tool supporting all the cost curve analysis described in this paper is available from the authors. Editors: Tom Faweett     

一种改进的心电信号预处理方法研究

心电信号是典型的强噪声下的非平稳微弱信号,减小噪声的干扰对心电信号的分析有着十分重要的意义,因此,有效的滤波方法一直是该领域学者关注的热点问题。本文在基于小波变换心电信号分析研究基础上,针对小波去噪时分解只作用于低频部分,从而忽略了高频区域中一部分有用信号的问题,提出了一种采用改进小波包理论实现心电信号去噪的方法,利用小波包在消除信号噪声方面具有更为精确的局部分析能力的特点,采用了‘db4’小波和＂最优基＂选择的方法,对心电信号进行消噪。以MIT-BIH心电数据库中心律失常数据仿真实验,得到了较理想的去噪效果。对比该方法与小波滤波去噪,发现基于小波包的心电信号去噪具有更优良的去噪性能。     

一种改进的群搜索优化方法

标准的群搜索优化(GSO)方法是一种适用于解决高维函数优化问题的群智能算法,且简单、高效,易于实现.为了进一步提高其收敛速度和精度,对该方法进行了改进.在保留其“发现者-追随者-游荡者”框架的同时,改进的GSO方法将最大下降方向策略引入发现者行为.在每轮迭代中,发现者不但按照自身方向进行搜索,同时也根据最大下降方向进行搜索.分别通过23个基准测试函数对2种优化方法进行测试,结果表明:改进的GSO方法优于标准群搜索方法.     

基于改进支持向量机的隐写分析方法

为了更有效地提高图像隐写分析的速度和正确检测率,提出了一种基于改进的支持向量机的隐写分析方法。采用Fridrich提出的多特征融合提取算法对图像进行特征提取,克服了单一特征不能很好描述图像差别的不足。然后提出了一种将最小二乘法与超球体一类支持向量机（HSOC-SVM）相结合的分类器——最小二乘超球一类支持向量机（LSHS-OCSVM）,并与目前广泛使用的FLD和非线性SVM分类器作对比实验。结果表明,方法是一种有效、高速的隐写分析方法。     

An improved method of substructure analysis

Substructure method is an established way of overcoming the difficulty of large dimensionality in analysing structures. An improved substructure analysis method suitable for structures like multi-storied buildings and towers is presented in this paper. The method is based on peculiar geometry of these structures which could be used for numbering the boundary joints for a substructure either in the beginning or towards the end. Extra advantage could also be taken of the substructures identical in terms of geometry and loading both or in terms of geometry alone. Special static condensation and substitutions routines are developed. The method is shown to be more efficient than any of the existing substructure analysis methods.     

基于改进的领导-跟随者编队算法研究

针对领导-跟随者算法的不足,提出了基于先驱者的多机器人编队控制方法.先驱者采用SLAM方法规划编队行进轨迹的关键点,领导者则根据关键点和环境信息,调整运动轨迹,并计算与跟随者间的相互距离和夹角,决定跟随者的实时位姿.当编队队形遇障不能通过时,采用基于逻辑层次的变换策略进行队形变换.同时在Player-Stage仿真平台上,对算法进行了比较和分析,实验结果表明,改进的领导-跟随者算法能更好的保证编队队形的完整性和编队系统的稳定性,控制效果较好.     

一种改进型i_p-i_q谐波检测法

介绍了基于瞬时无功功率理论的ip-iq法的基本原理,指出传统ip-iq法检测到的基波有功电流分量的幅值和相位因电网电压与其正序分量的相位差而存在误差;提出一种改进型ip-iq谐波检测法,该方法通过正序电压分量和移相算子相结合的方式获取同步旋转信号,减小了因三相电压不对称而导致的检测误差,同时引入电流平均值理论以取代低通滤波器,提高了谐波检测的动态响应能力。Matlab仿真结果表明,该检测法在三相电网电压不对称的情况下能准确检测到基波正序有功电流,且具有较好的动态响应特性。     

Rijndael算法中S_box及其改进方案

信息安全技术在网络银行,电子商务,电子现金,证书认证等领域发挥着极其重要的作用。而Rijndael作为高级加密标准,必成为信息安全的核心技术之一。本文介绍了Rijndael算法的原理与实现,研究了S_box的构造原理以及它的迭代输出周期性,指出S_box迭代输出周期过短的缺陷。提出了构造S_box的改进方案使其具有更好的代数性质以及更强的抗代数攻击的能力。     

改进的NeighLevel图像去噪方法

针对复小波变换在图像方向信息表征和NeighLevel算法刻画邻域相关性的局限性,提出了一种改进的图像去噪方法。首先,利用抗混叠轮廓波自由选择方向数的特点,能更好地提取图像边缘细节,克服了复小波方向性信息表达的不足;然后用变换域邻域小波系数之间的互信息量,改进NeighLevel方法对邻域信息的表达能力。理论分析和实验结果表明,与CWT-NeighLevel相比,在噪声方差等于30~60时,峰值信噪比提高了0.6%~7.0%,且在边缘特征方面保持了良好的视觉效果。     

An improved pseudospectral meshless radial point interpolation (PSMRPI) method for 3D wave equation with variable coefficients

In this paper, a pseudospectral meshless radial point interpolation (PSMRPI) technique is applied to the three-dimensional wave equation with variable coefficients subject to given appropriate initial and Dirichlet boundary conditions. The present method is a kind of combination of meshless methods and spectral collocation techniques. The point interpolation method along with the radial basis functions is used to construct the shape functions as the basis functions in the frame of the spectral collocation methods. These basis functions will have Kronecker delta function property, as well as unitary possession. In the proposed method, operational matrices of higher order derivatives are constructed and then applied. The merit of this innovative method is that, it does not require any kind of integration locally or globally over sub-domains, as it is essential in meshless methods based on Galerkin weak forms, such as element-free Galerkin and meshless local Petrov–Galerkin methods. Therefore, computational cost of PSMRPI method is low. Further, it is proved that the procedure is stable with respect to the time variable over some conditions on the 3D wave model, and the convergence of the technique is revealed. These latest claims are also shown in the numerical examples, which demonstrate that PSMRPI provides excellent rate of convergence.

     

一种基于改进支持向量机的入侵检测方法研究

提出基于粒子群优化（Particle Swarm Optimization,PSO）算法和支持向量机（Support Vector Machines,SVM）的入侵检测方法,为优化SVM性能,使用PSO的全局搜索特性寻找SVM的最优参数[C]和[σ];为避免PSO算法陷入局部最优,引入变异操作,找到最优参数组合后进行基于PSO_SVM入侵检测算法的训练和检测,解决了入侵检测系统准确度难题。仿真实验表明该方法的检测率为92.8%,误报率为6.911 9%,漏报率为9.708 7%,对KDDCUP竞赛的最佳结果有一定程度的提高,实验结果验证了该算法的有效性和可行性。     

An improved semianalytical method for sensitivity analysis

The semi-analytical method is conveniently used to obtain design sensitivities. However, it may have serious accuracy problems in shape design. In this study, an improved semianalytical method is presented for the accurate computation of shape design sensitivities. The method is based on approximating the flexibility matrix by means of von Neumann series. In numerical examples, two cases for which the standard semianalytical method fails are considered. It is demonstrated that the sensitivities can be obtained very accurately by the improved method proposed.     

一种改进扩展卡尔曼滤波新方法

针对现有的迭代扩展卡尔曼滤波（EIEKF）跟踪时估计精度较低这一不足,提出了一种改进扩展卡尔曼滤波（NIEKF）新方法。本文将迭代滤波理论引入到扩展卡尔曼滤波方法中,重复利用观测信息,采用经典的非线性非高斯模型进行仿真实验,给出了该方法与扩展卡尔曼滤波（EKF）、Unscented 卡尔曼滤波（UKF）、现有的迭代扩展卡尔曼滤波（EIEKF）的仿真结果,并分析了其跟踪性能和均方根误差。实验结果表明,改进扩展卡尔曼滤波（NIEKF）新方法具有更高的估计精度。     

OBB层次包围盒构造方法的改进

层次包围盒是碰撞检测的重要技术,可以提高碰撞检测精度并减少碰撞检测复杂度。首先介绍了OBB包围盒的计算方法、相交测试原理,以及OBB层次包围盒构造的一般过程。结合物体模型的层次结构,提出了基于模型层次结构的OBB层次包围盒的构造方法。实验结果证明改进后的方法加快了模块之间可以发生变化时其对应的OBB树的更新速度,使OBB方法适合由刚体模块构成的模型变形的情况。