Higher order eigensensitivity analysis of damped systems with repeated eigenvalues

A simplified method for the computation of first-, second- and higher-order derivatives of eigenvalues and eigenvectors associated with repeated eigenvalues is presented. Adjacent eigenvectors and orthonormal conditions are used to compose an algebraic equation. The algebraic equation which is developed can be used to compute derivatives of eigenvalues and eigenvectors simultaneously. Since the coefficient matrix in the proposed algebraic equation is non-singular, symmetric and based on N-space, it is numerically stable and very efficient compared to previous methods. To verify the efficiency of the proposed method, the finite element model of the cantilever beam and a mechanical system in the case of a non-proportionally damped system are considered.     

On a quadratic eigenproblem occurring in regularized total least squares

A computational approach for solving regularized total least squares problems via a sequence of quadratic eigenvalue problems has recently been proposed. Taking advantage of a variational characterization of real eigenvalues of nonlinear eigenproblems the existence of a real right-most eigenvalue for each quadratic eigenvalue problem in the sequence is proven. For large problems the approach is improved considerably utilizing information from the previous quadratic problems and early updates in a nonlinear Arnoldi method.     

自洽场方法中广义本征值方程求解及其C++程序设计

本文研究自洽场方法中广义本征值方程求解的算法,并设计相应的C 程序来实现该算法。首先对重叠矩阵进行分解,并将广义本征值方程化为标准的本征值方程,再利用Householder变换将上一步变换所得的矩阵化为对称三对角矩阵,进而用QL方法求解这个三对角矩阵的本征值和本征矢量,从而得到自洽场方法中广义本征值方程的本征值和本征矢量。     

A practical algorithm for the efficient computation of eigenvector sensitivities

Derivatives of eigenvalues and eigenvectors have become increasingly important in the development of modern numerical methods for areas such as structural design optimization, dynamic system identification and dynamic control, and the development of effective and efficient methods for the calculation of such derivatives has remained to be an active research area for several decades. In this paper, a practical algorithm has been developed for efficiently computing eigenvector derivatives of generalized symmetric eigenvalue problems. For eigenvector derivative of a separate mode, the computation only requires the knowledge of eigenvalue and eigenvector of the mode itself and an inverse of system matrix accounts for most computation cost involved. In the case of two close modes, the modal information of both modes is required and the eigenvector derivatives can be accurately determined simultaneously at minor additional computational cost. Further, the proposed method has been extended to the case of practical structural design where structural modifications are made locally and the eigenderivatives of the modes concerned before are still of interest. By combining the proposed algorithm together with the proposed inverse iteration technique and singular value decomposition theory, eigenproperties and their derivatives can be very efficiently computed. Numerical results from a practical finite element model have demonstrated the practicality of the proposed method. The proposed method can be easily incorporated into commercial finite element packages to improve the computational efficiency of eigenderivatives needed for practical applications.     

一种计算矩阵特征值特征向量的神经网络方法

当把Oja学习规则描述的连续型全反馈神经网络(Oja-N)用于求解矩阵特征值特征向量时,网络初始向量需位于单位超球面上,这给应用带来不便.由此,提出一种求解矩阵特征值特征向量的神经网络(1yNN)方法.在lyNN解析解基础上得到了以下结果:初始向量属于任意特征值对应特征向量张成的子空间,则网络平衡向量也将属于该空间;分析了lyNN收敛于矩阵最大特征值对应特征向量的初始向量取值条件;明确了lyNN收敛于矩阵不同特征值的特征子空间时,网络初始向量的最大取值空间;网络初始向量与已知特征向量垂直,则lyNN平衡解向量将垂直于该特征向量;证明了平衡解向量位于由非零初始向量确定的超球面上的结论.基于以上分析,设计了用lyNN求矩阵特征值特征向量的具体算法,实例演算验证了该算法的有效性.1yNN不出现有限溢,而基于Oja-N的方法在矩阵负定、初始向量位于单位超球面外时必出现有限溢,算法失效.与基于优化的方法相比,lyNN实现容易,计算量较小.     

On the efficient and accurate solution of the skew-symmetric eigenvalue problem: An arrangement of new and already known algorithmic formulations

Algorithms are presented to solve the special eigenvalue problem $\text{AZ}\text{=}\text{Zλ}$, where $\text{A}$ is skew-symmetric. The effective use of Householder's method, the bisection method and inverse iteration for solving the complete eigen-value problem are described in some detail. Simultaneous vector iteration is formulated for skew-symmetric matrices. The amount of work for the skew-symmetric Jacobi algorithm and the simultaneous vector iteration may be reduced by using the solution of a simplified eigenvalue problem. For Hermitian matrices also quadratic eigenvalue bounds for groups of eigenvalues and linear bounds for groups of eigenvectors are derived. The case where the set of calculated eigenvectors is not orthonormal is considered in some detail. In principle, the skew-symmetric eigenvalue problem may be easily transformed into a symmetric eigenvalue problem; but such a procedure has the following disadvantages: first, the results are in general less accurate, and, second, the eigenvectors which belong to well separated eigenvalues are not uniquely determined.     

用神经网络计算矩阵特征值与特征向量

该文研究用神经网格求解一般实对称矩阵的全部特征向量的问题。详细讨论了网络的平均态度合的结构并建立了平衡态集合的构造定理。通过求解简单的一维微分方程求出了网络的解析表达式。这一表达式是由对称矩阵的特征值与特征向量表达的、因而非常清晰利用解的解析表达式分析了网络的解的全局渐近行为。提出了用一些单位向量作为网络初始值计算对称矩阵的全部特征值与特征向量的具体算法。     

Efficient Rearrangement Algorithms for Shape Optimization on Elliptic Eigenvalue Problems

In this paper, several efficient rearrangement algorithms are proposed to find the optimal shape and topology for elliptic eigenvalue problems with inhomogeneous structures. The goal is to solve minimization and maximization of the k-th eigenvalue and maximization of spectrum ratios of the second order elliptic differential operator. Physically, these problems are motivated by the frequency control based on density distribution of vibrating membranes. The methods proposed are based on Rayleigh quotient formulation of eigenvalues and rearrangement algorithms which can handle topology changes automatically. Due to the efficient rearrangement strategy, the new proposed methods are more efficient than classical level set approaches based on shape and/or topological derivatives. Numerous numerical examples are provided to demonstrate the robustness and efficiency of new approach.     

一种求解复Hermite矩阵特征值的方法

介绍几种求解矩阵特征值和特征向量的经典算法及各自优缺点,通过理论推导,提出了一种性能稳健的方法,可以求解信号处理中常见的复Hermite阵.将对复Hermite矩阵求特征值和特征向量的问题转化为求解实对称阵的特征值和特征向量,而实对称阵的求解采用一种改进的三对角Householder法.最后把结果与Matlab仿真结果比较,可以看出该方法有很高的精确度.     

Partial solution of large symmetric generalized eigenvalue problems by nonlinear optimization of a modified Rayleigh quotient

A method is presented to solve numerically the lowest (or highest) eigenvalues and eigenvectors of the symmetric generalized eigenvalue problem. The technique proposed is iterative, does not transform the original matrices and yields eigencharacteristics in sequence, even for repeated eigenvalues. It is based on a nonlinear optimization of an unconstrained penalty function obtained from a generalization of the Rayleigh quotient. In addition, when the normality constraint is imposed, the eigenvectors are obtained by a sequence of solutions to linear equations, all with the same matrix. Examples demonstrate the validity of the method.     

极大代数意义下矩阵的特征值问题——一类离散事件动态系统运行周期的分析

在分析一类离散事件动态系统的运行周期及稳定性时,必须求解极大代数意义下矩阵的 特征值及特征向量,这一直被认为是十分困难和繁复的工作.本文给出了求任一方阵特征值 及特征向量的十分简单易行的方法以及有关的定理.     

Numerical solution of singular ODE eigenvalue problems in electronic structure computations

We put forward a new method for the solution of eigenvalue problems for (systems of) ordinary differential equations, where our main focus is on eigenvalue problems for singular Schrödinger equations arising for example in electronic structure computations. In most established standard methods, the generation of the starting values for the computation of eigenvalues of higher index is a critical issue. Our approach comprises two stages: First we generate rough approximations by a matrix method, which yields several eigenvalues and associated eigenfunctions simultaneously, albeit with moderate accuracy. In a second stage, these approximations are used as starting values for a collocation method which yields approximations of high accuracy efficiently due to an adaptive mesh selection strategy, and additionally provides reliable error estimates. We successfully apply our method to the solution of the quantum mechanical Kepler, Yukawa and the coupled ODE Stark problems.     

基于奇异值分解的图像压缩技术

为获得比较理想的图像压缩比和清晰的压缩后图像,使用了奇异值分解作为数据矩阵的压缩原理.详细解析了奇异值分解的原理及用奇异值分解压缩图像的原理.提出了按特征值个数占比阈值、按特征值之和占比阈值两种取特征值个数的方法.实验表明,特征值个数占比阈值在0.1时,图像清晰且压缩比达到5.99;特征值之和占比阈值在0.85时,图像清晰,对PNG格式图像压缩比达到7.89,对JPG格式图像压缩比达到5.92.从实验的个例来看,前1%的特征值表征了较多份量的数据特征,在征值个数占比阈值确定时,对PNG格式和对JPG格式图像压缩比相同;在特征值之和占比阈值确定时,对PNG格式相对比对JPG格式图像压缩比要高.认为按特征值之和占比阈值取特征值个数这种压缩方法更具普适性,可适用于Alpha通道冗余的情况,并可为大规模数量的图像压缩设定统一的特征值之和占比阈值.     

Discrete least-squares technique for eigenvalues. Part I: The one-dimensional case

A discrete least-squares technique for computing the eigenvalues of differential equations is presented. The eigenvalue approximations are obtained in two steps. Firstly, initial approximations of the desired eigenvalues are computed by solving a quadratic matrix eigenvalue problem resulting from the least-squares method applied to the equation under consideration. Secondly, these initial approximations, being of sufficient accuracy in some cases, are improved by using the Gauss-Newton method. Results from numerical experiments are reported that show great efficiency of the proposed technique in solving both regular and singular one-dimensional problems. The high flexibility of the technique enables one to use also the multidomain approach and the trial functions not satisfying any of the prescribed boundary conditions.     

Solving inverse eigenvalue problems by a projected Newton method

The inverse eigenvalue problem for symmetric matrices (IEP) can be formulated as a system of two matrix equations. For solving the system a variation of Newton's method is used which has been proposed by Fusco and Zecca [Calcolo XXIII (1986), pp. 285–303] for the simultaneous computation of eigenvalues and eigenvectors of a given symmetric matrix. An iteration step of this method consists of a Newton step followed by an orthonormalization with the consequence that each iterate satisfies one of the given equations. The method is proved to convergence locally quadratically to regular solutions. The algorithm and some numerical examples are presented. In addition, it is shown that the so-called Method III proposed by Friedland, Nocedal, and Overton [SIAM J. Numer. Anal., 24 (1987), pp. 634–667] for solving IEP may be constructed similarly to the method presented here.     

五对角矩阵的特征值反问题

本文讨论了一类由五个特征值和相应特征向量构造实对称五对角矩阵的特征值反问题.研究了解的存在性以及存在解的充分必要条件,而且给出了算法和数值例子.     

On preconditioned iterative methods for solving (A−λB)x=0

In this paper a preconditioned iterative method suitable for the solution of the generalised eigenvalue problem is presented. The proposed method in which no change in the structure of the original matrices occurs is suitable for the determination of the extreme eigenvalues and their corresponding eigenvectors of large sparse matrices derived from finite element/difference discretisation of partial differential equations. The new method when coupled with the conjugate gradient algorithm yields a powerful algorithm for this class of problems.     

认知无线电中基于特征值检测的频谱感知算法

频谱感知的任务在于利用感知节点(无线传感器或者认知用户)采集的数据判断频谱空洞是否存在。基于最大特征值检测(MED)和最小特征值检测(SED)的方法最近被应用到频谱感知当中。这两种算法在检测实际应用当中普遍存在的相关信号时表现出良好的检测性能。然而,MED和SED算法对应的判决门限求解非常复杂,从而限制了它们在实际的认知无线电频谱感知中的应用。该文利用取样协方差矩阵的所有特征值,提出了一种新的基于特征值检测(ESD)的算法。利用多元统计理论获得了相应的判决门限。ESD算法无需主信号和无线信道信息参与感知过程。与此同时,它保留了与MED和SED相同的计算复杂度。更重要的是ESD算法对应的判决门限可以通过一个简单的闭合表达式进行求解,其计算复杂度低。仿真结果验证了新算法的有效性。     

Poisson方程的Q_1~(rot)元和NF_1元特征值法

本文给出用三维非协调元的特征值方法求解一般的二阶椭圆边值问题的数值计算方法,从而验证了非协调元的收敛性的理论正确性及三维Q_1~(rot)元特征值误差渐进展开式的正确性.本文的数值实验表明:三维Q_1~(rot)元外推特征值下逼近准确特征值;三维NF_1元特征值和外推特征值都下逼近准确特征值;三维Q_1~(rot)元和三维NF_1元二网格离散方案特征值既下逼近准确特征值又上逼近准确特征值;三维Q_1~(rot)元比三维NF_1元有较好的数值表现.