无约束优化问题的对角稀疏拟牛顿法

对无约束优化问题提出了对角稀疏拟牛顿法,该算法采用了Armijo非精确线性搜索,并在每次迭代中利用对角矩阵近似拟牛顿法中的校正矩阵,使计算搜索方向的存贮量和工作量明显减少,为大型无约束优化问题的求解提供了新的思路．在通常的假设条件下,证明了算法的全局收敛性,线性收敛速度并分析了超线性收敛特征。数值实验表明算法比共轭梯度法有效,适于求解大型无约束优化问题．     

四种无约束优化算法的比较研究

从数值试验的角度 ,通过对 3个测试问题 (其中构造了一个规模大小可变的算例 )的求解 ,对共轭梯度法、BFGS拟牛顿法、DFP拟牛顿法和截断牛顿法进行比较研究 ,根据测试结果的分析 ,显示截断牛顿法在求解大规模优化问题时具有优势 ,从而为大规模寻优算法的研究提供了有益的借鉴 .     

正定可对称化矩阵与预对称迭代算法

１．问题的提出 我们引入正定可对称化矩阵定义的背景是为了研究求解二阶椭圆型非自共轭方程的离散迭代有效算法、这类方程的椭圆型是本质的分析性质。是由二阶项决定的,在离散方程中表现为正定性;非自共轭性则是由方程中的一阶项引起的,在相当广泛一类问题中可通过变量代换化为自共轭。因此,我们称这类问题为正定可对称化问题。 例１．高维二阶常系数椭圆型方程其中 Ａ为常系数正定对称（ｓ．ｐ．ｄ）阵, 为正交阵, Ｄ是对角元素为正的对角阵。 先作变量代换,通过演算,偏微分方程对于新变量变成这里进而令可将原非自共轭偏微分算子…     

一般线性或非线性约束下的共轭投影变尺度方法

梯度投影法是一类有效的约束最优化算法，在最优化领域中占有重要的地位．但是，梯度投影法所采用的投影是正交投影，不包含目标函数和约束函数的二阶导数信息·因而；收敛速度不太令人满意．本文介绍一种共轭投影概念，利用共轭投影构造了一般线性或非线性约束下的共轭投影变尺度算法，并证明了算法在一定条件下具有全局收敛性．由于算法中的共轭投影恰当地包含了目标函数和约束函数的二阶导数信息，因而收敛速度有希望加快．数值试验的结果表明算法是有效的．     

初始点任意的解非线性不等式约束优化问题的结合共轭梯度参数的超记忆梯度广义投影算法

本文利用广义投影矩阵，对求解无约束规划的超记忆梯度算法中的参数给出一种新的取值范围以保证得到目标函数的超记忆梯度广义投影下降方向，并与处理任意初始点的方法技巧结合建立求解非线性不等式约束优化问题的一个初始点任意的超记忆梯度广义投影算法，在较弱条件下证明了算法的收敛性．同时给出结合FR，PR，HS共轭梯度参数的超记忆梯度广义投影算法，从而将经典的共轭梯度法推广用于求解约束规划问题．数值例子表明算法是有效的．     

解带线性或非线性约束最优化问题的三项记忆梯度Rosen投影算法

利用Rosen投影矩阵，建立求解带线性或非线性不等式约束优化问题的三项记忆梯度Rosen投影下降算法，并证明了算法的收敛性．同时给出了结合FR，PR，HS共轭梯度参数的三项记忆梯度Rosen投影算法，从而将经典的共轭梯度法推广用于求解约束规划问题．数值例子表明算法是有效的。     

求解无约束优化问题的新谱共轭梯度法及其收敛性

强Wolfe条件不能保证标准CD共轭梯度法全局收敛．本文通过建立新的共轭参数,提出无约束优化问题的一个新谱共轭梯度法,该方法在精确线搜索下与标准CD共轭梯度法等价,在标准wolfe线搜索下具有下降性和全局收敛性．初步的数值实验结果表明新方法是有效的,适合于求解非线性无约束优化问题．     

拟牛顿法中的数值相关技术

处理病态无约束最优化问题不仅是其自身的需要,它也是解约束最优化问题、非线性方程组、偏微分方程边值问题的需要,本文提出的拟牛顿法中的数值相关技术,为利用拟牛顿法解决这类问题提供了一个手段。 众所周知,拟牛顿法是一类最有效的无约束最优化方法。不少拟牛顿公式具有校正     

一个具有n步二阶收敛速率的算法

本文考虑求解非线性方程组。从非线性ABS算法出发,建立了一类新算法。这类新算法具有更好的收敛性质;与求解无约束最优化的数值方法相对照,在某种意义上原非线性ABS算法对应于共轭梯度法,而本文的算法则对应于变度量法。     

冲击接触问题的一种双共振投影梯度算法

根据冲击接触计算模型所需满足的基本控制方程和非线性互补条件,应用非线性互补问题与约束优化的等价关系将非线性互补接触问题转变成一个非线性规划问题,系统地推导建立了冲击接触问题的一种双共轭投影梯度计算方法．增广Lagrange乘子法克服了罚函数要求减小迭代步长以达到计算稳定的限制,即使对于冲击接触问题亦可以采用较大迭代步长,在形成的与原互补问题等价的无约束规划模式下,应用双共轭投影梯度算法提高非线性搜索速度和计算效率．算法模型计算结果表明,所建立的双共轭投影梯度计算理论及方法是正确有效的．     

An inverse problem for the acoustics equations: A multilevel adaptive algorithm

A numerical solution to an inverse problem for the acoustic equations using an optimization method for a stratified medium is presented. With the distribution of an acoustic wave field on the medium’s surface, the 1D distributions of medium’s density, as well as the velocity and absorption coefficient of the acoustic wave, are determined. Absorption in a Voigt body model is considered. The conjugate gradients and the Newton method are used for minimization. To increase the efficiency of the numerical method, a multilevel adaptive algorithm is proposed. The algorithm is based on a division of the whole procedure of solving the inverse problem into a series of consecutive levels. Each level is characterized by the number of parameters to be determined at the level. In moving from one level to another, the number of parameters changes adaptively according to the functional minimized and the convergence rate. The minimization parameters are chosen as illustrated by results of solving the inverse problem in a spectral domain, where the desired quantities are presented as Chebyshev polynomial series and minimization is carried out with respect to the coefficients of these series. The method is compared in efficiency with a nonadaptive method. The optimal parameters of the multilevel method are chosen. It is shown that the multilevel algorithm offers several advantages over the one without partitioning into levels. The algorithm produces primarily a more accurate solution to the inverse problem.     

AN ADAPTIVE MULTI-SCALE CONJUGATE GRADIENT METHOD FOR DISTRIBUTED PARAMETER ESTIMATION OF 2-D WAVE EQUATION

An adaptive multi-scale conjugate gradient method for distributed parameter estimations (or inverse problems) of wave equation is presented. The identification of the coefficients of wave equations in two dimensions is considered. First, the conjugate gradient method for optimization is adopted to solve the inverse problems. Second, the idea of multi-scale inversion and the necessary conditions that the optimal solution should be the fixed point of multi-scale inversion method is considered. An adaptive multi-scale inversion method for the inoerse problem is developed in conjunction with the conjugate gradient method. Finally, some numerical results are shown to indicate the robustness and effectiveness of our method.     

Recovery of an interface from boundary measurement in an elliptic differential equation

We study the inverse problem of recovering an interior interface from a boundary measurement in an elliptic boundary value problem arising from a semiconductor transistor model. We set up a nonlinear least-squares formulation for solving the inverse problem, and establish the necessary derivatives with respect to the interface. We then propose both the Gauss–Newton iterative method and the conjugate gradient method for the least-squares problem, and present implementation of these methods using integral equations.     

The method of fundamental solutions for the Oseen steady‐state viscous flow past obstacles of known or unknown shapes

In this paper, the steady‐state Oseen viscous flow equations past a known or unknown obstacle are solved numerically using the method of fundamental solutions (MFS), which is free of meshes, singularities, and numerical integrations. The direct problem is linear and well‐posed, whereas the inverse problem is nonlinear and ill‐posed. For the direct problem, the MFS computations of the fluid flow characteristics (velocity, pressure, drag, and lift coefficients) are in very good agreement with the previously published results obtained using other methods for the Oseen flow past circular and elliptic cylinders, as well as past two circular cylinders. In the inverse obstacle problem the boundary data and the internal measurement of the fluid velocity are minimized using the MATLAB^© optimization toolbox lsqnonlin routine. Regularization was found necessary in the case the measured data are contaminated with noise. Numerical results show accurate and stable reconstructions of various star‐shaped obstacles of circular, bean, or peanut cross‐section.     

Splitting-based conjugate gradient method for a multi-dimensional linear inverse heat conduction problem

In this paper we consider a multi-dimensional inverse heat conduction problem with time-dependent coefficients in a box, which is well-known to be severely ill-posed, by a variational method. The gradient of the functional to be minimized is obtained by the aid of an adjoint problem, and the conjugate gradient method with a stopping rule is then applied to this ill-posed optimization problem. To enhance the stability and the accuracy of the numerical solution to the problem, we apply this scheme to the discretized inverse problem rather than to the continuous one. The difficulties with large dimensions of discretized problems are overcome by a splitting method which only requires the solution of easy-to-solve one-dimensional problems. The numerical results provided by our method are very good and the techniques seem to be very promising.     

刚性目标形状反演的一种非线性最优化方法

发展了从声散射场的远场分布的信息来再现声刚性目标形状反问题的一种非线性最优化方法,它是通过独立地求解一个不适定的线性系统和一个适定的非线性最小化问题来实现的。对反问题的非线性和不适定性的这种分离式数值处理,使所建立方法的数值实现是非常容易和快速的,因为在确定声刚性障碍物形状的非线性最优化步中,只需求解一个只有一个未知函数的小规模的最小平方问题。该方法的另一个特别的性质是,只需要远场分布的一个Fourier系数,即可对未知的刚性目标作物形设别。进而提出了数值实现该方法的一种两步调整迭代算法。对具有各种形状的二维刚性障碍物的数值试验保证了本算法是有效和实用的。     

Inverse Problem for a Class of Two-Dimensional Diffusion Equations with Piecewise Constant Coefficients

In this paper, we consider an inverse problem for a class of two-dimensional diffusion equations with piecewise constant coefficients. This problem is studied using an explicit formula for the relevant spectral measures and an asymptotic expansion of the solution of the diffusion equations. A numerical method that reduces the inverse problem to a sequence of nonlinear least-square problems is proposed and tested on synthetic data.     

The Inverse Problem for Numerical Determination of the Nonlinear Right-Hand Side in a Parabolic Equation

The article considers the inverse problem of determining the nonlinear right-hand side of a quasi-linear parabolic equation and proves a uniqueness theorem. A method is proposed for numerical solution of the inverse problem based on parametric representation of the sought coefficient. The inverse problem thus reduces to finding the error-minimizing vector of unknown coefficients of the parametric representation of the sought function.     

Reconstruction of a source domain from boundary measurements

Inverse and ill-posed problems which consist of reconstructing the unknown support of a source from a single pair of exterior boundary Cauchy data are investigated. The underlying dependent variable, e.g. potential, temperature or pressure, may satisfy the Laplace, Poisson, Helmholtz or modified Helmholtz partial differential equations (PDEs). For constant coefficients, the solutions of these elliptic PDEs are sought as linear combinations of explicitly available fundamental solutions (free-space Greens functions), as in the method of fundamental solutions (MFS). Prior to this application of the MFS, the free-term inhomogeneity represented by the intensity of the source is removed by the method of particular solutions. The resulting transmission problem then recasts as that of determining the interface between composite materials. In order to ensure a unique solution, the unknown source domain is assumed to be star-shaped. This in turn enables its boundary to be parametrized by the radial coordinate, as a function of the polar or, spherical angles. The problem is nonlinear and the numerical solution which minimizes the gap between the measured and the computed data is achieved using the Matlab toolbox routine lsqnonlin which is designed to minimize a sum of squares starting from an initial guess and with no gradient required to be supplied by the user. Simple bounds on the variables can also be prescribed. Since the inverse problem is still ill-posed with respect to small errors in the data and possibly additional ill-conditioning introduced by the spectral feature of the MFS approximation, the least-squares functional which is minimized needs to be augmented with regularizing penalty terms on the MFS coefficients and on the radial function for a stable estimation of these couple of unknowns. Thorough numerical investigations are undertaken for retrieving regular and irregular shapes of the source support from both exact and noisy input data.     

TV‐like regularization for backward parabolic problems

This paper investigates an inverse problem for parabolic equations backward in time, which is solved by total‐variation‐like (TV‐like, in abbreviation) regularization method with cost function ∥u_x∥². The existence, uniqueness and stability estimate for the regularization problem are deduced in the linear case. For numerical illustration, the variational adjoint method, which presents a simple method to derive the gradient of the optimization functional, is introduced to reconstruct the unknown initial condition for both linear and nonlinear parabolic equations. The conjugate gradient method is used to iteratively search for the optimal approximation. Numerical results validate the feasibility and effectiveness of the proposed algorithm. Copyright © 2016 John Wiley & Sons, Ltd.