A nonmonotone line search filter method with reduced Hessian updating for nonlinear optimization

This paper proposes a nonmonotone line search filter method with reduced Hessian updating for solving nonlinear equality constrained optimization. In order to deal with large scale problems, a reduced Hessian matrix is approximated by BFGS updates. The new method assures global convergence without using a merit function. By Lagrangian function in the filter and nonmonotone scheme, the authors prove that the method can overcome Maratos effect without using second order correction step so that the locally superlinear convergence is achieved. The primary numerical experiments are reported to show effectiveness of the proposed algorithm.     

A filter secant method with nonmonotone line search for equality constrained optimization

<正> This paper proposes a filter secant method with nonmonotone line search for non-linearequality constrained optimization.The Hessian of the Lagrangian is approximated using the BFGSsecant update.This new method has more flexibility for the acceptance of the trial step and requires lesscomputational costs compared with the monotone one.The global and local convergence of the proposedmethod are given under some reasonable conditions.Further,two-step Q-superlinear convergence rateis established by introducing second order correction step.The numerical experiments are reported toshow the effectiveness of the proposed algorithm.     

A line search filter inexact SQP method for nonlinear equality constrained optimization

This paper proposes an inexact SQP method in association with line search filter technique for solving nonlinear equality constrained optimization.For large-scale applications,it is expensive to get an exact search direction,and hence the authors use an inexact method that finds an approximate solution satisfying some appropriate conditions.The global convergence of the proposed algorithm is established by using line search filter technique.The second-order correction step is used to overcome the Maratos effect,while the line search filter inexact SQP method has q-superlinear local convergence rate.Finally,the results of numerical experiments indicate that the proposed method is efficient for the given test problems.     

A Dwindling Filter Line Search Algorithm for Nonlinear Equality Constrained Optimization

This paper proposes a dwindling filter line search algorithm for nonlinear equality constrained optimization. A dwindling filter, which is a modification of the traditional filter, is employed in the algorithm. The envelope of the dwindling filter becomes thinner and thinner as the step size approaches zero. This new algorithm has more flexibility for the acceptance of the trial step and requires less computational costs compared with traditional filter algorithm. The global and local convergence of the proposed algorithm are given under some reasonable conditions. The numerical experiments are reported to show the effectiveness of the dwindling filter algorithm.     

AN ADAPTIVE TRUST REGION METHOD FOR EQUALITY CONSTRAINED OPTIMIZATION

In this paper, a trust region method for equality constrained optlmization based on nondiferentiable exact penalty is proposed. In this algorithin, the trail step is characterized by computation of its normal component being separated from computation of its tangential component, i.e., only the tangential component of the trail step is constrained by trust radius while the normal component and trail step itself have no constraints. The other main characteristic of the algorithm is the decision of trust region radius. Here, the decision of trust region radius uses the information of the gradient of objective function and reduced Hessian. However, Maratos effect will occur when we use the nondifferentiable exact penalty function as the merit function. In order to obtain the superlinear convergence of the algorithm, we use the twice order correction technique. Because of the speciality of the adaptive trust region method, we use twice order correction when p= 0 (the definition is as in Section 2) and this is different from the traditional trust region methods for equality constrained opthnization. So the computation of the algorithm in this paper is reduced. What is more, we can prove that the algorithm is globally and superlinearly convergent.     

A Superlinearly Convergent Combined PhaseⅠ-PhaseⅡ Subfeasible Method

ＡＳｕｐｅｒｌｉｎｅａｒｌｙＣｏｎｖｅｒｇｅｎｔＣｏｍｂｉｎｅｄＰｈａｓｅⅠ－ＰｈａｓｅⅡＳｕｂｆｅａｓｉｂｌｅＭｅｔｈｏｄＪＩＡＮＪｉｎｂａｏ（ＭａｔｈｅｍａｔｉｃｓａｎｄＩｎｆｏｒｍａｔｉｏｎＳｃｉｅｎｃｅＤｅｐａｒｔｍｅｎｔｏｆＧｕａｎｇｘｉＵ...     

An affine scaling derivative-free trust region method with interior backtracking technique for bounded-constrained nonlinear programming

This paper proposes an arlene scaling derivative-free trust region method with interior backtracking technique for bounded-constrained nonlinear programming. This method is designed to get a stationary point for such a problem with polynomial interpolation models instead of the objective function in trust region subproblem. Combined with both trust region strategy and line search technique, at each iteration, the affine scaling derivative-free trust region subproblem generates a backtracking direction in order to obtain a new accepted interior feasible step. Global convergence and fast local convergence properties are established under some reasonable conditions. Some numerical results are also given to show the effectiveness of the proposed algorithm.     

A family of the local convergence of the improved secant methods for nonlinear equality constrained optimization subject to bounds on variables

This paper studies a family of the local convergence of the improved secant methods for solving the nonlinear equality constrained optimization subject to bounds on variables. The Hessian of the Lagrangian is approximated using the DFP or the BFGS secant updates. The improved secant methods are used to generate a search direction. Combining with a suitable step size, each iterate switches to trial step of strict interior feasibility. When the Hessian is only positive definite in an affine null subspace, one shows that the algorithms generate the sequences converging q-linearly and two-step q-superlinearly. Furthermore, under some suitable assumptions, some sequences generated by the algorithms converge locally one-step q-superlinearly. Finally, some numerical results are presented to illustrate the effectiveness of the proposed algorithms.     

A nonmonotone filter line search technique for the MBFGS method in unconstrained optimization

This paper presents a new nonmonotone filter line search technique in association with the MBFGS method for solving unconstrained minimization. The filter method, which is traditionally used for constrained nonlinear programming （NLP）, is extended to solve unconstrained NLP by converting the latter to an equality constrained minimization. The nonmonotone idea is employed to the filter method so that the restoration phrase, a common feature of most filter methods, is not needed. The global convergence and fast local convergence rate of the proposed algorithm are established under some reasonable conditions. The results of numerical experiments indicate that the proposed method is efficient,     

A new <Emphasis Type="Italic">ɛ</Emphasis>-generalized projection method of strongly sub-feasible directions for inequality constrained optimization

In this paper, the nonlinear optimization problems with inequality constraints are discussed. Combining the ideas of the strongly sub-feasible directions method and the ɛ-generalized projection technique, a new algorithm starting with an arbitrary initial iteration point for the discussed problems is presented. At each iteration, the search direction is generated by a new ɛ-generalized projection explicit formula, and the step length is yielded by a new Armijo line search. Under some necessary assumptions, not only the algorithm possesses global and strong convergence, but also the iterative points always get into the feasible set after finite iterations. Finally, some preliminary numerical results are reported.     

A Strong Subfeasible Directions Algorithm with Superlinear Convergence

ＡＳｔｒｏｎｇＳｕｂｆｅａｓｉｂｌｅＤｉｒｅｃｔｉｏｎｓＡｌｇｏｒｉｔｈｍｗｉｔｈＳｕｐｅｒｌｉｎｅａｒＣｏｎｖｅｒｇｅｎｃｅＪＩＡＮＪｉｎｂａｏ（Ｄｅｐｔ．ｏｆＭａｔｈ．ａｎｄＩｎｆｏｒｍａｔｉｏｎＳｃｉｅｎｃｅ，ＧｕａｎｇｘｉＵｎｉｖｅｒｓｉｔｙ...     

具有动态步长和发现概率的布谷鸟搜索算法

为了进一步改善算法搜索过程中存在的求解精度偏低、收敛速度缓慢等现象,提出具有动态步长和发现概率的布谷鸟搜索算法。该算法通过引入步长调整因子动态约束每一代种群的莱维移动步长,使算法的莱维飞行机制具有自适应性。在发现概率上,使用具有均匀分布和F分布特性的随机惯性权重,改变发现概率的固定取值,加强种群的多样性,保持算法全局搜索、局部探索之间的平衡状态。通过实验证明,所提算法具有良好的可行性,其寻优结果、收敛速度均有提高。     

A new descent memory gradient method and its global convergence

In this article, a new descent memory gradient method without restarts is proposed for solving large scale unconstrained optimization problems. The method has the following attractive properties: 1) The search direction is always a sufficiently descent direction at every iteration without the line search used; 2) The search direction always satisfies the angle property, which is independent of the convexity of the objective function. Under mild conditions, the authors prove that the proposed method has global convergence, and its convergence rate is also investigated. The numerical results show that the new descent memory method is efficient for the given test problems.     

约束优化问题的一个超线性收敛的广义梯度投影法

对线性约束的非线性优化问题提出了一个新的广义梯度投影法，该算法我们采用了非精确线性搜索，并在每次迭代运算中运用了广义投影矩阵和变尺度方法的思想确定其搜索方向。在通常的假设条件下，证明了算法的整体收敛性和超线性收敛速度。     

An Active-Set Projected Trust Region Algorithm for Box Constrained Optimization Problems

An active-set projected trust region algorithm is proposed for box constrained optimization problems, where the given algorithm is designed by three steps. First, the projected gradient direction which normally has better numerical performance is introduced. Second, the projected trust region direction that often possesses good convergence is defined, where the matrix of trust region subproblem is updated by limited memory strategy. Third, in order to get both good numerical performance and convergence, the authors define the final search which is the convex combination of the projected gradient direction and the projected trust region direction. Under suitable conditions, the global convergence of the given algorithm is established. Numerical results show that the presented method is competitive to other similar methods.     

混合方法优化的自适应引力搜索算法

针对引力搜索算法存在的易早熟收敛、易陷入局部最优、搜索精度有待提高等缺陷,提出一种混合方法优化的自适应引力搜索算法（gravitational search algorithm,GSA）。首先利用Sobol序列初始化种群,增强算法全局搜索能力;其次引入Hamming贴进度计算种群成熟度,判断种群是否早熟;然后引入Logistic混沌对种群作混沌搜索,变异已陷入局部最优的粒子位置;最后基于早熟收敛判断因子改进引力系数,并为粒子位置公式添加收缩因子,促使种群加快脱离局部最优。对9个不同类型的基准测试函数做仿真实验,结果表明新算法能有效改善种群的早熟问题,具备更好的寻优性能。     

OFDM系统中相位噪声的自适应补偿技术及性能仿真

相位噪声对正交频分复用OFDM系统的性能有着关键性影响,其产生的通用相角错误CPE会使信道失真系数逐符号旋转,因而必须采用逐符号更新信道估计值的方法。提出了一种基于判决反馈的低通滤波算法来克服CPE所产生的信道旋转并提高信道估计的性能。该方法可通过导频子载波估计CPE所产生的信道旋转,并将解调后子载波的判决值通过低通滤波的方式降低信道估计的方差。文中采用了归纳法对该算法进行了理论分析,分析结果和仿真均证明了该算法的有效性,以及在计算量,收敛性方面的优势。     

基于混合算法的MIMO雷达正交多相码设计

提出了一种基于遗传算法和禁忌搜索算法的多输入多输出雷达正交多相码波形设计方法,并将其用于类零相关正交多相码的设计。将禁忌搜索算法引入遗传算法,充分考虑遗传算法的全局收敛性和禁忌搜索算法的局部收敛性。为给禁忌搜索算法一个好的初始解,先用遗传算法优化到一定程度再用禁忌搜索算法,即遗传算法迭代多次,禁忌搜索算法迭代一次。采用最优保存策略来避免最优解丢失,使发射信号的自相关峰值旁瓣和互相关峰进一步降低,提高主副比。仿真结果验证了所提方法的可行性和有效性。     

基于洛伦兹函数的变步长凸组合最小均方算法

为进一步减小收敛速率与稳态误差之间的矛盾,改善自适应滤波算法,利用改进的Lorentzian函数提出了一种新的变步长凸组合最小均方(new variable step-size convex combination of least mean square,NVS-CLMS)算法,该算法既有效提高了收敛速率又具备很好的抗干扰能力。同时,为了克服CLMS算法停滞等待的弊端,采用了瞬时转移结构;另外,在参数的迭代公式中使用sign函数进行优化以降低运算量。仿真结果证明该算法与CLMS、VS-CLMS相比,在不同的仿真环境中均能表现出良好的均方特性和跟踪特性。     

基于相关矩阵对角化和遗传算法的盲源分离法

提出了一种新的基于相关矩阵对角化的代价函数作为衡量输出信号独立性的测度。为了扩大搜索空间,降低各信源之间的互相关性,将代价函数进行了非线性变换。还提出了利用实数编码的遗传算法对代价函数进行最优化搜索,以克服传统梯度搜索方法容易陷入局部收敛的缺点。此方法不仅适用于平稳或非平稳信号,而且还可用于瞬时或卷积混和模型的盲源分离问题。仿真实验表明,该算法具有快速收敛性能和高精确度等优点,能够大大提高分离后的输出信噪比。