基于split-and-conquer和非参数向前选择法的变量选择

变量选择是统计学界研究的重要课题之一.当处理高维数据时,一些常用的变量选择方法大多比较耗时,因此提出了一种既能筛选出高维数据中的变量又能节省时间的方法:基于split-and-conquer的非参数向前选择法.首先使用split-and-conquer方法将数据进行拆分,然后使用B样条函数逼近的非参数向前选择法进行研究.实验结果表明:基于split-and-conquer的非参数向前选择法可以较好地将变量选择出来,并且节省了大量时间.     

A non-overlapping domain decomposition dual reciprocity method for solving the forward–backward heat equation in two-dimension

We apply a boundary element dual reciprocity method (DRBEM) to the numerical solution of the forward–backward heat equation in a two-dimensional case. The method is employed for the spatial variable via the fundamental solution of the Laplace equation and the Crank–Nicolson finite difference scheme is utilized to treat the time variable. The physical domain is divided into two non-overlapping subdomains resulting in two standard forward and backward parabolic equations. The subproblems are then treated by the underlying method assuming a virtual boundary in the interface and starting with an initial approximate solution on this boundary followed by updating the solution by an iterative procedure. In addition, we show that the time discrete scheme is unconditionally stable and convergent using the energy method. Furthermore, some computational aspects will be suggested to efficiently deal with the formulation of the proposed method. Finally, two forward–backward problems, for which the exact solution is available, will be numerically solved for two different domains to demonstrate the efficiency of the proposed approach.     

Near‐forward/high‐pressure‐backward Raman study of Zn1 − xBexSe (x ~ 0.5) – evidence for percolation behavior of the long (Zn―Se) bond

The apparently universal 1‐bond → 2‐mode percolation behavior in the Raman spectra of zincblende semiconductor alloys is generally observed for the short bond only, and not for the long one. In this work we perform a combined high‐pressure‐backward/near‐forward Raman study of the leading percolation‐type (Zn,Be)Se alloy (~50 at.% Be), which exhibits a distinct percolation doublet in the spectral range of its short Be―Se bond, in search of a Zn―Se analogue. The high‐pressure‐backward insight is not conclusive per se, but clarifies the perspective behind the near‐forward Raman study. The latter reveals an unique Zn―Se phonon–polariton. Its fair contour modeling depending on the scattering angle is achieved within the linear dielectric approach, based on ellipsometry measurement of the ZnBeSe refractive index. Somewhat surprisingly this reveals that the phonon–polariton in question is a ‘fractional’ one in that it carries only half of the available Zn―Se oscillator strength, as ideally expected in case of a BeSe‐like bimodal Raman behavior of the long Zn―Se bond. Copyright © 2015 John Wiley & Sons, Ltd.     

Proximal-gradient algorithms for fractional programming

In this paper, we propose two proximal-gradient algorithms for fractional programming problems in real Hilbert spaces, where the numerator is a proper, convex and lower semicontinuous function and the denominator is a smooth function, either concave or convex. In the iterative schemes, we perform a proximal step with respect to the nonsmooth numerator and a gradient step with respect to the smooth denominator. The algorithm in case of a concave denominator has the particularity that it generates sequences which approach both the (global) optimal solutions set and the optimal objective value of the underlying fractional programming problem. In case of a convex denominator the numerical scheme approaches the set of critical points of the objective function, provided the latter satisfies the Kurdyka-?ojasiewicz property.     

THE THEORETICAL COST OF SEQUENTIAL AND PARALLEL ALGORITHMS FOR SOLVING LINEAR SYSTEMS OF EQUATIONS

ＴＨＥ　ＴＨＥＯＲＥＴＩＣＡＬ　ＣＯＳＴ　ＯＦ　ＳＥＱＵＥＮＴＩＡＬ　ＡＮＤ　ＰＡＲＡＬＬＥＬ　ＡＬＧＯＲＩＴＨＭＳ　ＦＯＲ　ＳＯＬＶＩＮＧ　ＬＩＮＥＡＲ　ＳＹＳＴＥＭＳ　ＯＦ　ＥＱＵＡＴＩＯＮＳＳａｌｍａｎＨ．Ａｂｂａｓ（ＲｅｃｅｉｖｅｄＭａｙ２４...     

A New Scheme for High‐Intensity Laser‐Driven Electron Acceleration in a Plasma

We propose a new approach to high‐intensity relativistic laser‐driven electron acceleration in a plasma. Here, we demonstrate that a plasma wave generated by a stimulated forward‐scattering of an incident laser pulse can be in the longest acceleration phase with injected relativistic beam electrons. This is why the plasma wave has the maximum amplification coefficient which is determined by the acceleration time and the breakdown (overturn) electric field in which the acceleration of the injected beam electrons occurs. We must note that for the longest acceleration phase the relativity of the injected beam electrons plays a crucial role in our scheme. We estimate qualitatively the acceleration parameters of relativistic electrons in the field of a plasma wave generated at the stimulated forward‐scattering of a high‐intensity laser pulse in a plasma. (© 2015 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)     

地表反照率对短波红外探测大气CO2的影响

在卫星短波红外遥感二氧化碳的过程中, 表征地表特征的地表反照率是影响探测精度的重要参数之一. 针对温室气体二氧化碳高精度探测的需求, 本文研究了地表反照率对正演模拟光谱和反演近地面二氧化碳平均柱浓度XCO₂的影响. 模拟计算结果显示, 地表反照率数值增大时, 观察的光谱强度也相应增大, 并且在O₂-A波段造成的光谱差异比1.6 μm波段高出一个数量级, 即地表反照率在O₂-A波段的影响比较大. 选取了两个不同地表类型的实际观测光谱, 仅改变O₂-A波段和1.6 μm波段地表反照率数值, 得出草地点在O₂-A波段地表反照率达到0.25的误差时, 会给XCO₂的反演结果造成大于1%的相对误差, 而1.6 μm波段的地表反照率变化对XCO₂的反演结果造成的误差可以忽略不计, 说明了地表反照率在反演XCO₂过程中的重要性主要来自对O₂-A波段的影响. 此研究表明了地表反照率在卫星遥感温室气体过程中的重要性, 为提高遥感探测二氧化碳的精度提供了重要的理论依据和指导.     

Impacts of TV and radio advertisements on the dynamics of an infectious disease: A modeling study

TV and radio advertisements are widely acknowledged as important interventions in raising issues of public health care and play promising role to control the infection through propagating awareness among the individuals. In this paper, a nonlinear susceptible‐infected‐susceptible (SIS) model is proposed and analyzed to see the impacts of TV and radio advertisements on the spread of influenza epidemic. In the model formulation, it is assumed that the susceptible individuals contract infection through the direct contact with infected individuals. The information regarding the protection against the disease is propagated via TV and radio advertisements, and their growth rates are assumed to be proportional to the fraction of infected individuals. However, the growth rate of TV advertisements decreases with the increase in number of aware individuals. The information broadcasted through TV and radio advertisements induces behavioral changes among the susceptible individuals, and they form an isolated aware class. The epidemiological feasible equilibria, their stability properties, and direction of bifurcation are discussed. The expression for modified basic reproduction number is obtained. The model analysis shows that the dissemination rate of awareness among susceptible individuals due to TV and radio advertisements and baseline number of TV and radio advertisements have potential to reduce the epidemic peak and, thus, control the spread of infection. Further, the analytical findings are well supported through numerical simulation.     

A mean reverting process for pricing treasury bills and futures contracts

An analytically tractable, discrete-time single-factor model is developed for valuing treasury bills and futures contracts. It uses a multiplicative binomial foward process that creates neither negative nor implausibly large positive interest factors, and which can incorporate different possible degrees of mean reversion. The paper derives explicit formulae for bill prices, futures prices, their conditional variances and risk premia in a setting that relates the evolution of the term structure more closely to both model and data than do other similar works. In contrast to other term-structure constrained models, this paper emphasizes that in a one-factor model the martingale probabilities cannot be treated independently of the perturbation functions. The paper's empirical methods also differ from the customary approaches. Instead of comparing differences between model-predicted and observed prices, the paper applies ARCH methodology to test model-predicted ratios of conditional variances to risk premia. Our tests find influences exogenous to the model, but these factors do not seem capable of being explained with two-factor models using only interest rates.     

激光尾流场的2.5维粒子模拟

 用2D3V PIC粒子模拟方法分析了超短脉冲超强激光在稀薄等离子体中激发尾流场的产生过程及电子在尾流场中的加速过程。“前向Raman散射”使得激光脉冲沿传播方向拉长，脉冲的尾部变陡，它导致静电场的相速度和饱和时超热电子的最大动能明显减小，也使得激发尾流场的最佳脉冲宽度变小。