A SUPERLINEARLY CONVERGENT SPLITTING FEASIBLE SEQUENTIAL QUADRATIC OPTIMIZATION METHOD FOR TWO-BLOCK LARGE-SCALE SMOOTH OPTIMIZATION

This paper discusses the two-block large-scale nonconvex optimization problem with general linear constraints.Based on the ideas of splitting and sequential quadratic optimization(SQO),a new feasible descent method for the discussed problem is proposed.First,we consider the problem of quadratic optimal(QO)approximation associated with the current feasible iteration point,and we split the QO into two small-scale QOs which can be solved in parallel.Second,a feasible descent direction for the problem is obtained and a new SQO-type method is proposed,namely,splitting feasible SQO(SF-SQO)method.Moreover,under suitable conditions,we analyse the global convergence,strong convergence and rate of superlinear convergence of the SF-SQO method.Finally,preliminary numerical experiments regarding the economic dispatch of a power system are carried out,and these show that the SF-SQO method is promising.

Finite Nuclei in a Relativistic mean—Field Theory with quadratic Couplings

A quadratic scalar and vector coupling model proposed recently has been applied to finite nuclei.The calculated results are compared with those of the derivative scalar coupling (DSC) model and the nonlinear Walecka model The results show that the spin-orbit splittings are improved considerably by quadratic couplings in contrast to the DSC model However,the binding energy per nucleon,rms charge radius,as well as the spin-orbit splittings in the quadratic model are still small compared with those given by the nonlinear Walecka model and the experimental data.

Modulus-Based Matrix Splitting Iteration Methods for a Class of Stochastic Linear Complementarity Problem

For the expected value formulation of stochastic linear complementarity problem, we establish modulus-based matrix splitting iteration methods. The convergence of the new methods is discussed when the coefficient matrix is a positive definite matrix or a positive semi-definite matrix, respectively. The advantages of the new methods are that they can solve the large scale stochastic linear complementarity problem, and spend less computational time. Numerical results show that the new methods are efficient and suitable for solving the large scale problems.

High-Order Decoupled and Bound Preserving Local Discontinuous Galerkin Methods for a Class of Chemotaxis Models

In this paper,we explore bound preserving and high-order accurate local discontinuous Galerkin(LDG)schemes to solve a class of chemotaxis models,including the classical Keller-Segel(KS)model and two other density-dependent problems.We use the convex splitting method,the variant energy quadratization method,and the scalar auxiliary variable method coupled with the LDG method to construct first-order temporal accurate schemes based on the gradient flow structure of the models.These semi-implicit schemes are decoupled,energy stable,and can be extended to high accuracy schemes using the semi-implicit spectral deferred correction method.Many bound preserving DG discretizations are only worked on explicit time integration methods and are difficult to get high-order accuracy.To overcome these difficulties,we use the Lagrange multipliers to enforce the implicit or semi-implicit LDG schemes to satisfy the bound constraints at each time step.This bound preserving limiter results in the Karush-Kuhn-Tucker condition,which can be solved by an efficient active set semi-smooth Newton method.Various numerical experiments illustrate the high-order accuracy and the effect of bound preserving.

A Large-update Interior-point Algorithm for Convex Quadratic Semi-definite Optimization Based on a New Kernel Function

In this paper, we present a large-update interior-point algorithm for convex quadratic semi-definite optimization based on a new kernel function. The proposed function is strongly convex. It is not self-regular function and also the usual logarithmic function. The goal of this paper is to investigate such a kernel function and show that the algorithm has favorable complexity bound in terms of the elegant analytic properties of the kernel function. The complexity bound is shown to be O（√n（logn）2 log e/n）. This bound is better than that by the classical primal-dual interior-point methods based on logarithmic barrier function and in optimization fields. Some computational results recent kernel functions introduced by some authors have been provided.

POSITIVE DEFINITE AND SEMI-DEFINITE SPLITTING METHODS FOR NON-HERMITIAN POSITIVE DEFINITE LINEAR SYSTEMS

In this paper, we further generalize the technique for constructing the normal （or pos- itive definite） and skew-Hermitian splitting iteration method for solving large sparse non- Hermitian positive definite system of linear equations. By introducing a new splitting, we establish a class of efficient iteration methods, called positive definite and semi-definite splitting （PPS） methods, and prove that the sequence produced by the PPS method con- verges unconditionally to the unique solution of the system. Moreover, we propose two kinds of typical practical choices of the PPS method and study the upper bound of the spectral radius of the iteration matrix. In addition, we show the optimal parameters such that the spectral radius achieves the minimum under certain conditions. Finally, some numerical examples are given to demonstrate the effectiveness of the considered methods.

基于非局部先验的深度压缩感知图像重构网络

传统的基于迭代的压缩感知(CS)图像重构算法易于集成图像先验信息,但存在性能不足、计算复杂度高等缺点。基于深度学习的图像重构算法重构性能通常优于传统的重构算法,并且具有更低的重构计算成本。因此,为了设计出一种更有效利用先验信息的深度学习图像重构算法,该文提出基于非局部先验的深度压缩感知图像重构网络。首先,将稀疏性和非局部先验相结合建立压缩感知图像重构模型,然后通过半二次方分裂法将模型分解为3个子问题,每一个子问题的求解都在深度学习的框架下展开,最后联合建立端到端的可训练的图像重构模型。仿真实验表明,在测试的采样率与数据集下该文所提算法的峰值信噪比与当前主流的重构算法SCSNet相比平均提升了0.18 dB,与CSNet算法相比平均提升了约1.59 dB,与ISTA-Net+算法相比平均提升了约2.09 dB。

Regular Splitting and Potential Reduction Method for Solving Quadratic Programming Problem with Box Constraints

Presents a regular splitting and potential reduction method for solving a quadratic programming problem with box constraints. Discussion on the regular splitting and potential reduction algorithm; Complexity analysis of the algorithm; Analysis of the complexity bound on obtaining an approximate solution.

A SPLITTING METHOD FOR QUADRATIC PROGRAMMING PROBLEM

A matrix splitting method is presented for minimizing a quadratic programming (QP) problem, and a general algorithm is designed to solve the QP problem and generates a sequence of iterative points. We prove that the sequence generated by the algorithm converges to the optimal solution and has an R-linear rate of convergence if the QP problem is strictly convex and nondegenerate, and that every accumulation point of the sequence generated by the general algorithm is a KKT point of the original problem under the hypothesis that the value of the objective function is bounded below on the constrained region, and that the sequence converges to a KKT point if the problem is nondegenerate and the constrained region is bounded.

基于稀疏先验的合成孔径光学系统的图像复原方法

由于通光面积减小以及系统共相误差等原因,合成孔径光学系统的成像会出现降质模糊。针对该问题,提出一种基于L_(0)范数的稀疏先验图像复原方法。对合成孔径系统的退化过程建模,通过统计分析与数学验证,发现退化后图像的暗通道值变成非极小值。基于暗通道理论提出图像复原算法模型,将暗通道和梯度先验以L_(0)范数的形式作为正则化项进行约束,光学系统阵列结构下计算的点扩散函数作为初始模糊核;最后,使用半二次分裂法迭代求解最终估计的清晰图像。在仿真实验中,不同活塞误差下和多场景下方法均能够提升图像的对比度、还原细节部分,迭代5次复原后的平均峰值信噪比分别达到23.13 dB和23.79 dB,平均结构相似度分别达到0.77和0.80,优于维纳滤波与Richardson-Lucy方法。在合成孔径相机实际拍摄图像的复原中,本文方法仍然能够有效还原退化图像。

基于l_(p)范数的分裂算法在稀疏角CT图像重建中的应用

稀疏角CT图像重建是热点研究问题。根据lp范数具有良好的稀疏性及小波变换能够提取图像的全局和局部特征的性质,提出一个基于l_(p)范数及小波变换的稀疏角CT图像重建模型。利用半二次分裂算法对模型进行求解,并证明了该模型的全局收敛性。数值实验结果表明,该算法可以很好地抑制伪影,得到清晰的边缘,并且在投影角度较少的情况下也能取得较好的重建效果。

大规模非凸不可分优化问题的分裂序列二次规划算法

该文研究了目标函数和约束函数带不可分结构的大规模非凸优化问题,提出了一个新的分裂序列二次规划算法.首先,借助分裂算法思想将传统二次规划(QP)子问题的增广拉格朗日问题分解为两个小规模QP子问题,通过求解小规模QP子问题产生改进的搜索方向.其次,以增广拉格朗日函数作效益函数,通过Armijo线搜索产生下一个迭代点.在较为温和的条件下,获得新算法的全局收敛性.最后,对该算法进行了数值实验,验证了算法的有效性。

基于多种先验结合的图像盲去模糊方法

针对传统图像盲去模糊方法中部分细节复原不准确的问题,提出了一种结合暗通道先验、梯度先验以及强度先验的图像盲去模糊方法。该方法利用清晰图像本身的稀疏性以及边缘信息的稀疏性,结合清晰图像暗通道的稀疏性,在不同权重下对清晰图像以及模糊核进行约束。通过交替迭代的方式更新潜影和模糊核,以估计出的最精细的模糊核结合非盲去模糊方法最终得到去模糊图像。实验结果表明,所提方法对图像去模糊质量有所提升,有效地恢复了图像的一些细节。

非结构网格FVTD中二次函数重构技术用于电磁计算

提出了一种适用于非结构网格下格子中心型时域有限体积方法的基于最小平方法的二次函数重构技术以取代以往采用的线性重构技术,包括插值模板的选择和重构函数的确定。通过对典型的电磁场本征值问题及散射问题的计算验证,表明了文中方法较以往方法大大改善了算法的耗散性,提高精度的同时可以显著地减少CPU时间和存储量。

采用子网流量组合优化的网络流量分割方法

针对当前基于仿真的流量分割方法存在处理开销过大且仅能保证流层面流量完整性的问题,提出了一种组合优化的网络流量分割方法.将流量经分割后得到的数据包总延迟次数作为优化目标,由此网络流量分割问题转化为二次分配优化问题;用子网流量作为基本单元进行分配,可在减小问题求解规模的同时保证子网流量的完整性.利用真实网络流量对所提方法的分割效果进行了验证,并与基于数据包和流的轮询流量分割方法进行了比较,结果表明,所提方法能够达到基于流的轮询流量分割方法的性能,数据包延迟较之不采用优化的基于子网流量轮询方法显著减小,性能提高近10%.

基于水平集方法的均布式柔性机构的拓扑优化设计

提出一种利用水平集方法进行均布式柔性机构设计的新方法。根据水平集边界表达方法中具有几何信息的特点,将图像分析中的二次能量函数引入到水平集模型中,以控制柔性机构拓扑优化设计结果的几何尺寸,得到等宽带状均布的柔性机构,较好地解决了传统柔性机构拓扑优化中容易出现单点铰链问题。应用半隐式的加性分裂算子(AOS)算法求解水平集方程,松弛了逆风格式中CFL(Courant-Frie drichs-Lewy)条件对时间步长的限制,提高了求解效率。通过一个典型的二维算例来验证方法的有效性。

裂区设计二因素试验结果的二次回归分析法

本文根据最小二乘原理提出裂区设计二因素试验结果的二次回归分析法，包括建 立试验指标与试验因素间的二元二次回归方程和显著性检验。此法开辟了分析裂 区试验结果的新途径，能从试验结果中提取新的重要信息，从而提高了试验结果 的利用率，最后举例说明．

二元裂解算子交替方向乘子法的核极限学习机

凸优化形式的核极限学习机(KELM)具有较高的分类准确率,但用迭代法训练凸优化核极限学习机要较传统核极限学习机的解线性方程法花费更长时间。针对此问题,该文提出一种2元裂解算子交替方向乘子法(BSADMM-KELM)来提高凸优化核极限学习机的训练速度。首先引入2元裂解算子,将求核极限学习机最优解的过程分裂为两个中间算子的优化过程,再通过中间算子的迭代计算而得到原问题的最优解。在22个UCI数据集上所提算法的训练时间较有效集法平均快29倍,较内点法平均快4倍,分类精度亦优于传统的核极限学习机;在大规模数据集上该文算法的训练时间优于传统核极限学习机。

余代数上的同调二次型

文章从简单的余模范畴中的正合列入手,一方面分析了遗传artin余代数,利用长度公式定义了同调二次型及正定二次型,并讨论同调二次型满足正定性所需条件及相应结果;另一方面研究遗传anin余代数的有限表示型的一些必要条件,并逐渐与同调二次型的正定性相联系;最后构造遗传artin余代数上的同调二次型正定性与有限表示型之间的充分必要关系．

基于双L_0稀疏先验的图像运动去模糊

目前图像的运动去模糊方法在处理较复杂的运动模糊时难以得到理想的效果,其原因之一是这些方法普遍只考虑图像梯度的稀疏性,忽略了模糊核的稀疏性。针对这一不足提出一种新的双L_0正则约束的运动模糊去除方法,该方法同时对自然图像梯度和模糊核使用L_0正则约束,结合半正定二次分裂最小化的方法进行求解优化,实现自然模糊图像梯度和模糊核均稀疏下的模糊核估计,并进一步使用L_(0.5)超拉普拉斯正则约束项恢复最终图像。实验发现,该方法可以较好地去除单幅图像较复杂的运动模糊,更好地克服了估计的模糊核中存在的噪点和错误,得到较现有方法更加理想去模糊效果。