一种改进的SVM算法

当训练集的规模很大特别是支持向量很多时,支持向量机的学习过程需要占用大量的内存,寻优速度非常缓慢,这给实际应用带来了很大的麻烦。文献[4]提出了一种针对大规模样本集的学习策略,该方法虽大幅降低了学习的代价,但存在着一个致命的弱点:如果初始样本集选择不当,SVM的分类精度将得不到保障。基于此,本文引入了“最远邻”,对文献[4]中算法进行了改进。实验表明,采用这种改进的算法不仅保留了文献[4]方法的优点,而且这样获得的分类器的分类精度完全可以与直接通过大规模样本集训练得到的分类器的分类精度相媲美,甚至更优。     

Sampling Bandpass Signals

In a recent paper [1], Brown examined the sampling of a real finiteenergy bandpass signal having an (angular) bandwidth ? (in radians per second) at the theoretically minimum (average) rate of ?/? samples per second. Following Grace and Pitt's [2] quadrature sampling, a particular case of Kohlenberg's second-order sampling [3, 4], Brown has proved the feasibility of a separate interpolation of the in-phase and quadrature components of the signal when ?o = k?/2 (Brown's condition), where ?o is the center (angular) frequency of the signal and k is an arbitrary positive integer. Here the problem is reconsidered from a general point of view, introducing, under Brown's condition, an interpolation formula which includes that of Grace and Pitt and which extends a theorem of Populis [5-7]. We indicate the necessary and sufficient conditions to obtain a separate interpolation, offering closed formulas to obtain the interpolation functions. We also discuss the minimum oversampling rate needed whenBrown's condition is not verified.     

叶轮机械任意旋成面叶栅气动杂交命题矩函数型变分有限元解法

本文在文[1]、[2]提供的矩函数变分原理基础上建立了任意旋成面叶栅A类、B类杂交气动命题的变分有限元解法。文中用有限元法对相应的变分原理进行离散,所得非线性代数方程组迭代求解。由解出的矩函数计算流场速度、压力与密度,最后通过积分获得叶片未知边界形状。本文给出的计算结果显示了杂交命题方法的有效性和灵活性,它可以直接用于叶片改型与设计工作。     

Comments on "on Probability of Error in Decision Feedback Equalizer"

The above-mentioned paper [1] derives bounds on the error probability of a decision feedback equalizer (DFE) which are claimed to be significantly tighter than the earlier upper bound of [2]. Unfortunately, [1] makes an unsupportable assumption which reddars the conclusions incorrect. We show that valid bounds can be obtained by the methods of [1], but that for the numerical example of [2] , the resulting upper bound is never tighter than, and sometimes significantly weaker than, that derived in [2]. Similarly, a lower bound derived by the method of [1] is shown for the same numerical example to be very close to the error-free past lower bound previously known.     

Track-While-Scan Algorithm in a Clutter Environment

A track-while scan (TWS) algorithm is developed for targets in a clutter environment. The problem has been studied using only the position measurements [1, 5-8], but the simulation results have not been satisfactory. Modern processing techniques (FFT processor) ) in air traffic control and surveillance radar receivers provide both position and radial velocity. The radial velocity measurement may be conveniently used in the target-track correlation process, which will reduce the association ambiguity in the clutter environment. t. In the clear environment the algorithm using the position and radial velocity measurements has been treated in [3, 4]. A TWS algorithm, using both position and radial velocity measurements for targets in a clutter environment, is presented here. The algorithm obtained is nonlinear and adaptive. In order to evaluate the improvement due to radial velocity measurement a simulation has been performed on a digital computer. The algorithm was run with and without radial velocity measurements to compare its performances. An improvement was noted especially when the target path included an accelerated portion.     

镍基单晶合金叶片振动特性分析

利用通用有限元程序Marc分析了单晶叶片的振动特性,考虑了取向、温度和转速的影响。计算结果表明,[111]取向的固有频率大于[001]取向的,取向对高阶固有频率的影响比低阶的大;温度升高,固有频率下降;随着转速增加,固有频率也在增加。转速对[001]取向固有频率的影响要比[111]取向的大。在工作状态下,转速对低阶频率的影响大,对高阶频率的影响则比较小。     

跨声速无粘流绕振动翼型流动的数值模拟

本文从Euler方程出发,利用LU-ADI格式对绕二元振动翼型跨声速流场进行了数值模拟。文章中给出了绕NACA0012和NACA 64A 010翼型不同状态的几组数值结果,在同某些实验和其它文献结果基本吻合的前提下,研究了振动对翼型表面压力、激波位置的非定常效应以及其它振动参数对流场非定常效应的影响。     

镍基单晶合金涡轮叶片榫头裂纹尖端应力场及扩展趋势

采用率相关晶体滑移有限元程序,考虑单晶材料晶体取向的影响,对镍基单晶合金涡轮叶片榫头裂纹特性进行有限元分析.分别计算了榫头裂纹在各向同性条件以及{001},{011},{111}三种不同晶体取向下裂纹尖端的Mises应力分布,并判断了裂纹的扩展趋势.结果表明:镍基单晶合金涡轮叶片在{001}[110],{011}[110],{111}[110]取向下的裂纹尖端均存在着明显的应力集中和较大的应力梯度,应力的最大值存在于裂纹尖端;{001},{011},{111}三种不同晶体取向的裂纹尖端的扩展开裂角分别为45°,54.7°,90°,说明镍基单晶合金涡轮叶片的裂纹扩展趋势受晶体取向的影响较大.      

Efficient fault tolerant estimation using the IMM methodology

Space systems are characterized by a low-intensity process noise resulting from uncertain forces and moments. In many cases, their scalar measurement channels can be assumed to be independent, with one-dimensional internal dynamics. The nominal operation of these systems can be severely damaged by faults in the sensors. A natural method that can be used to yield fault tolerant estimates of such systems is the interacting multiple model (IMM) filtering algorithm, which is known to provide very accurate results. However, having been derived for a general class of systems with switching parameters, the IMM filter does not utilize the independence of the measurement errors in different channels, nor does it exploit the fact that the process noise is of low intensity. Thus, the implementation of the IMM in this case is computationally expensive. A new estimation technique is proposed herein, that explicitly utilizes the aforementioned properties. In the resulting estimation scheme separate measurement channels are handled separately, thus reducing the computational complexity. It is shown that, whereas the IMM complexity is exponential in the number of fault-prone measurements, the complexity of the proposed technique is polynomial. A simulation study involving spacecraft attitude estimation is carried out. This study shows that the proposed technique closely approximates the full-blown IMM algorithm, while requiring only a modest fraction of the computational cost.     

B-样条函数的一种快速递推法

 本文提出了常用的等距结点B-样条函数的一种快速递推法及递推计算格式和程序设计。阐明了此法具有计算简便、递推快速和更为直观的特点。它为B-样条函数的数学分析和推广应用提供了方便。此法将易于为一般工程技术人员掌握和运用。     

On the Sensitivity of a Fixed-Point Smoothing Algorithm

This paper presents a simple approach to the derivation of sensitivity measures of smoothing algorithms. It is proposed that the known sensitivity results of the Kalman filtering algorithm be utilized along with the state augmentation approach for this purpose. The sensitivity measures so obtained are easier to compute than the ones available in the literature. It is then shown that the fixed-point smoothing algorithm, derived recently by Biswas and Mahalanabis [1], [2], is less sensitive to model parameter variations than the algorithm studied by Griffin and Sage [7], [8]. The case of a satellite tracking problem is presented by way of illustrating the results.     

关于计入剪切效应的板弯曲杂交模式的若干问题

本文从Reissner变分原理出发,构造一计入剪切效应的9自由度板弯曲杂交元。通过离散型Kirchhoff约束的实施,把厚板公式直接推广于薄板计算而不出现解的“闭锁”现象。文中提出的能量调整原理有效地解决了厚板元素中的弯、剪匹配问题,从而实现了厚薄板分析的统一性。文章最后给出计算实例。     

Comments on "Angular Spectra in Radar Applications"

In [1], we proposed using "data adaptive" spectrum estimation techniques (e.g., maximum likelihood and maximum entropy) for radar problems such as separating low angle returns from ground reflections. W. White suggests [2] that these techniques will be of little value due to the correlation between the various received signals. In this rebuttal, we present additional conceptual arguments, simulation results and field data which support the utility of the techniques set forth in [1].     

用一种新的极点配置方程确定双输入多输出系统的控制律

 本文导出了双输入多输出系统的极点配置方程。它们的形式和通常所用的极点配置方程形式不同。证明了方程的一个性质,利用这个性质可以较好地控制各个特征值配置的精度。算例表明,用这种方法配置的极点的精度比用“秩一控制器”方法有较大改善。     

含孔层合板考虑边界效应的拉伸强度分析

1.引言 早在1974年和1975年Whitney与Nuismer就提出了预测含孔层合板拉伸强度的点应力准则和平均应力准则,他们假定的某一特征长度是与材料性能有关系。1977年Karlak只就准各向同性碳/环氧层合板的两种铺层顺序在特征长度中引进了考虑铺层顺序影响的     

A Reverse Monte Carlo Study of H+D Lyman Alpha Absorption from QSO Spectra

A new method based on a Reverse Monte Carlo [RMC] technique and aimed at the inverse problem in the analysis of interstellar (intergalactic) absorption lines is presented. The line formation process in chaotic media with a finite correlation length (l > 0) of the stochastic velocity field (mesoturbulence) is considered. This generalizes the standard assumption of completely uncorrelated bulk motions (l 0) in the microturbulent approximation which is used for the data analysis up-to-now. It is shown that the RMC method allows to estimate from an observed spectrum the proper physical parameters of the absorbing gas and simultaneously an appropriate structure of the velocity field parallel to the line-of-sight.The application to the analysis of the H+D Ly profile is demonstrated using Burles and Tytler [B&T] data for QSO 1009+2956 where the DI Ly line is seen at za = 2.504.The results obtained favor a low D/H ratio in this absorption system, although our upper limit for the hydrogen isotopic ratio of about 4.5 × 10-5 is slightly higher than that of B&T (D/H = 30 _-0.5 ^+0.6 × 10^-5). We also show that the D/H and N(HI) values are, in general, correlated, i.e. the derived D-abundance may be badly dependent on the assumed hydrogen column density. The corresponding confidence regions for an arbitrary and a fixed stochastic velocity field distribution are calculated.     

复合材料层板等强最优设计方法

 本文对复合材料无矩对称层板的最优设计,提出了一种概念直观、方法简便、收敛迅速、效果较好的等强最优设计方法。给出了几个算例的结果,并与文献[1、2]的方法作了对比,结果表明,本文的方法具有显著的优越性。     

基于混合遗传算法求解飞机平衡状态

通过分析飞机平衡状态的基本特性和现有配平算法存在的问题,提出了一种求解不同类型飞机多种平衡状态的通用方法。该方法与飞机模型结构无关,只利用配平约束进行计算,算法采用与单纯形相结合的混合遗传算法,提高了搜索速度与精度。仿真结果表明这种方法具有较好的收敛性与稳定性。     

基于改进鸽群算法的气动捕获轨道优化

针对火星探测器的气动捕获轨道，提出了一种改进的鸽群算法对气动捕获轨道进行优化。首先，考虑成功进行气动捕获所要求的终端约束和过程约束，根据从捕获轨道进行轨道转移进入目标轨道所需的速度增量，提出了进行捕获轨道优化的最优性能指标。然后，针对原始鸽群算法存在的一些不足，提出一种改进算法，并对改进算法的参数取值进行了分析。最后，基于大气内飞行的动力学方程，将气动捕获轨道优化问题转化为多参数优化问题，利用所提出的改进鸽群算法对气动捕获轨道进行优化，并通过仿真实例验证该算法的有效性。     

Three-dimensional trajectory estimation from image position and velocity

A recursive algorithm for estimating the three-dimensional trajectory and structure of a moving rigid object in an image sequence has been previously developed by Broida, Chandrashekhar, and Chellappa [1990]. Since then, steady advances have occurred in the calculation of optical flow. This work improves 3D motion trajectory and structure estimation by incorporating optical flow into the estimation framework. The new solution combines optical flow and feature point measurements and determines their statistical relationship. The feasibility of a hybrid feature point/optical flow algorithm, demonstrated through detailed simulation on synthetic and real image sequences, significantly lowers bias and mean squared error in trajectory estimation over the feature-based approach.