Almost sure convergence of classification procedures using Hermite series density estimates

A multidimensional classification procedure is examined derived from the multiple Hermite series estimate of probability density functions. Conditions for the almost sure convergence of the integrated square error for the estimate are presented and the rate of the convergence is studied. The probability of misclassification, conditioned on a learning sequence of length n, is shown to converge to the Bayes risk almost surely as rapidly as $\text{O}\text{(n}{}^{\mathrm{<mtext>?1</mtext><mtext>2+\delta </mtext>}}\text{)}$, δ positive.     

A bivariate C Clough-Tocher scheme

A Clough-Tocher like interpolation scheme is described for values of position, gradient and hessian at scattered points in two variables. The domain is assumed to have been triangulated. The interpolant has local support, is globally twice differentiable, piecewise polynomial, and reproduces polynomials of degree up to three exactly.     

Multi-User Ultra-Wide Band Communication System Based on Modified Gegenbauer and Hermite Functions

In this paper a multi-user communication system based on ultra-wide band (UWB) technology is studied. UWB uses very short pulses, so that the spectrum of the emitted signals may spread over several GHz. In order to implement multi-user communication, the emitted signal must be modulated. One modulation scheme for UWB communications is to use analog waveforms to modulate the data. In this study, orthogonal waveforms called modified Gegenbauer functions are introduced as basis functions for the pulse shape and compared to the previously proposed modified Hermite functions. It is shown that Gegenbauer functions offer better performance for multi-user UWB communications. Fouzia Elbahhar was born in 1975. She received the M.S. and Ph.D. degrees from the University of Valenciennes (France) in 2000 and 2004, respectively. She is actually employed as Research Asistant at this university. She is especially involved in Ultra Wide Band technology. Her primary interest is in systems dedicated to land transportation for communications, like vehicle to vehicle telecommunication. Atika Rivenq-Menhaj was born in 1970. She received her Diploma of Engineering and the M.S. degree in 1993 and then her Ph.D. degree in 1996, from the University of Valenciennes (France). She actually is Assistant Professor in electronics at this university. Her primary interest is in signal processing applied to intelligent transportation systems and telecommunication systems. Jean-Michel Rouvaen was born in 1947. He received his M.S. degree in 1968 and his Ph.D. degree in 1971, from the University of Valenciennes (France). He is now Professor of electronics at ENSIAME, an engineering school of university of Valenciennes. His primary interests are in signal processing, nonlinear phenomena and telecommunications.     

Current distribution in a parallel set of conducting strips

The current distribution in a parallel set of thin conducting sheets due to an external applied source is investigated. All sheets are placed in one plane. The source, and all excited fields, are time-harmonic. The frequency is low enough to allow for an electro quasi-static approximation (neglecting the displacement current). The conducting sheets are infinitely long and the current is uniform in the longitudinal direction of the sheets. The sheets have a thin rectangular cross-section, so thin that the current can be assumed uniform in the thickness-direction. Hence, the current distribution only depends on the transverse coordinate. Due to the mutual induction between the sheets, the current distribution over the width of the cross-section becomes non-uniform: it accumulates at the edges of the sheets. It is especially this so-called edge-effect, and its dependence on the applied frequency and the distances between the sheets, that is the aim of this investigation. From the Maxwell equations, a set of integral equations for the current distribution in the sheets is derived. These integral equations are solved, as far as possible by analytical means, by writing the current distribution in each sheet as a series of Legendre polynomials. The general method is worked out for N (N 1) sheets, but explicit results are presented for N=1 and 3. It turns out that the edge-effect becomes stronger for increasing frequencies. For this solution, only a very restricted number of Legendre polynomials are needed.     

Solution of the two-dimensional wave equation by using wave polynomials

The paper demonstrates a specific power-series-expansion technique to solve approximately the two-dimensional wave equation. As solving functions (Trefftz functions) so-called wave polynomials are used. The presented method is useful for a finite body of certain shape geometry. Recurrent formulas for the wave polynomials and their derivatives are obtained in the Cartesian and polar coordinate system. The accuracy of the method is discussed and some examples are shown.     

Short communication: Genetic variation of saturated fatty acids in Holsteins in the Walloon region of Belgium

Random regression test-day models using Legendre polynomials are commonly used for the estimation of genetic parameters and genetic evaluation for test-day milk production traits. However, some researchers have reported that these models present some undesirable properties such as the overestimation of variances at the edges of lactation. Describing genetic variation of saturated fatty acids expressed in milk fat might require the testing of different models. Therefore, 3 different functions were used and compared to take into account the lactation curve: (1) Legendre polynomials with the same order as currently applied for genetic model for production traits; 2) linear splines with 10 knots; and 3) linear splines with the same 10 knots reduced to 3 parameters. The criteria used were Akaike's information and Bayesian information criteria, percentage square biases, and log-likelihood function. These criteria indentified Legendre polynomials and linear splines with 10 knots reduced to 3 parameters models as the most useful. Reducing more complex models using eigenvalues seemed appealing because the resulting models are less time demanding and can reduce convergence difficulties, because convergence properties also seemed to be improved. Finally, the results showed that the reduced spline model was very similar to the Legendre polynomials model.     

基于最小二乘法的正交多项式级数在谐波估计中的应用

利用基于最小二乘法的正交多项式级数来逼近谐波的概率密度函数并估计其95%不超过概率值。根据电力谐波测量样本统计样本的阶矩(原点矩);对于多变量或多矢量求和,首先建立求和模型,在统计出分量阶矩的基础上通过递归方法计算和的阶矩。基于最小二乘法的思想,确定逼近的正交多项式级数展开中的待定系数。这种方法不要求谐波的幅值和相位具有某种特定分布,适用范围广。最后用实例证明了该方法的有效性。     

Rijndael算法中S盒优化测试方法研究

文章首先说明了Rijndael算法中S盒的四种测试方法,包括差分、线性、雪崩和布尔表达式测试。使用有限域中30个最高次幂为8的不可约多项式利用MATLAB构造不同S盒,然后对这些S盒进行优化测试,发现它们在差分和线性测试上性能相似,在雪崩和布尔表达式测试方面的差别主要在稳定性和分布均匀性上,从这两方面得到了最优的S盒。     

非高斯随机过程的短期极值估计:复合Hermite模型

Hermite模型自20世纪80年代后期开始被广泛应用于非高斯随机过程的短期极值估计。当随机过程的非高斯性很强时,尤其是偏度很大时,常用的3阶Hermite模型不足以表征出极值分布的尾端特征。工程中,样本统计矩的不确定性使得更高阶的Hermite模型不宜使用。基于此,该文提出了同时基于中心矩与线性矩的复合Hermite模型,有效地将Hermite模型由3阶拓展到4阶。该文以对数正态模型作为非线性系统的研究对象,对比分析了在解析条件下和在使用蒙特卡洛模拟获得样本数据条件下,各类Hermite模型与传统的Gumbel法以及平均条件穿越率（ACER）法用于极值分析的表现。结果表明,对于大偏度强非高斯随机过程的极值预测,复合Hermite模型具有更好的精确度和鲁棒性。