• 论文 • 上一篇    下一篇

不等式约束优化一个基于滤子思想的广义梯度投影算法

简金宝1,2, 马鹏飞3, 徐庆娟4   

  1. 1. 玉林师范学院数学与信息科学学院, 广西玉林 537000;
    2. 广西大学数学与信息科学学院, 南宁 530004;
    3. 上海大学理学院, 上海 200444;
    4. 广西师范学院数学与信息科学学院, 南宁 530001
  • 收稿日期:2012-11-16 出版日期:2013-05-15 发布日期:2013-05-13
  • 通讯作者: 简金宝
  • 基金资助:

    国家自然科学基金(11271086)、广西自然科学基金(2011GXNSFD018022)和广西高校人才小高地建设创新团队资助计划.

简金宝, 马鹏飞, 徐庆娟. 不等式约束优化一个基于滤子思想的广义梯度投影算法[J]. 计算数学, 2013, 35(2): 205-214.

Jian Jinbao, Ma Pengfei, Xu Qingjuan. A GENERALIZED GRADIENT PROJECTION METHOD BASED ON THE IDEA OF FILTER FOR INEQUALITY CONSTRAINED OPTIMIZATION[J]. Mathematica Numerica Sinica, 2013, 35(2): 205-214.

A GENERALIZED GRADIENT PROJECTION METHOD BASED ON THE IDEA OF FILTER FOR INEQUALITY CONSTRAINED OPTIMIZATION

Jian Jinbao1,2, Ma Pengfei3, Xu Qingjuan4   

  1. 1. School of Mathematics and Informatics Science, Yulin Normal University, Yulin 537000, Guangxi, China;
    2. School of Mathematics and Informatics Science, Guangxi University, Nanning 530004, China;
    3. Shanghai University College of Science, Shanghai 200444, China;
    4. Mathematics and Information Science,Teachers Education University, Nanning 530001, Guangxi, China
  • Received:2012-11-16 Online:2013-05-15 Published:2013-05-13
讨论非线性不等式约束优化问题, 借鉴于滤子算法思想,提出了一个新型广义梯度投影算法.该方法既不使用罚函数又无真正意义下的滤子.每次迭代通过一个简单的显式广义投影法产生搜索方向,步长由目标函数值或者约束违反度函数值充分下降的Armijo型线搜索产生.算法的主要特点是: 不需要迭代序列的有界性假设;不需要传统滤子算法所必需的可行恢复阶段;使用了ε积极约束集减小计算量.在合适的假设条件下算法具有全局收敛性, 最后对算法进行了初步的数值实验.
In this paper, optimization problems with nonlinear inequality constraints are discussed. Based on the idea of the filter algorithm, a new generalized gradient projection algorithm is proposed. The proposed method uses neither a penalty function, nor a strict filter. At each iteration of the proposed algorithm, the search direction is yield by just on explicit generalized gradient projection. The step-size is selected such that either the value of the objective function or the measure of the constraint violations is sufficiently reduced by a Armijo line search technique. The main properties of the proposed algorithm as follows: don’t need to assume the boundness of iteration sequence; don’t need any restoration phase which is necessary for filter methods; the scale and the computation cost are further decreased by using the ε-active set. The algorithm is globally convergent under suitable assumptions. Finally, some elementary numerical experiments are reported.

MR(2010)主题分类: 

()
[1] Rosen J B. The Gradient Projection Method of Nonlinear Programming, Part I, Linear Constraints[J]. SIAM J Appl Math, 1960, 8(1): 181-21.

[2] Rosen J B. The gradient projection method for nonlinear programming part I linear constraints[J].SIAM J Appl Math, 1961, 9(4): 514-532.

[3] 堵丁柱, 孙捷. 一个新的梯度投影方法[J]. 计算数学, 1983, 4: 378-386.

[4] 章祥荪. 关于梯度投影法的收敛性证明[J]. 应用数学学报,1985, 1(1): 125-128.

[5] Jian Jinbao. A new feasible descent algorithm combining SQP with generalized projection foroptimization problems without strict complementarity[J]. Appl Math Comput, 2005, 162(3):1065-1081.

[6] 赖炎连, 简金宝.初始点任意的一个非线性优化的广义梯度投影法[J]. 系统科学与数学, 1995,15(4): 374-380.

[7] Jian Jinbao, Ma Guodong, Guo Chuanhao. A new ε-generalized projection method of stronglysub-feasible directions for inequality constrained optimization[J]. Journal of System Science andComplex, 2011, 24(1): 604–618.

[8] Fletcher R, Leyffer S. Nonlinear programming without a penalty function[J]. Math Program, 2002,91: 239-269.

[9] Fletcher R, Gould N, Leyffer S, et al. Global convergence of a trust-region SQP-filter algorithmfor general nonlinear programming[J]. SIAM J Optim, 2002, 13(3): 635-659.

[10] Ulbrich S. On the superlinear local convergence of a filter-SQP methods[J]. Math Program, 2004,100: 217-245.

[11] Wäechter A, Biegler. Line search filter methods for nonlinear programming : motivation andglobal convergence[J]. SIAM J Optim, 2005, 16(1): 1-31.

[12] Wäechter A, Biegler. Line search filter methods for nonlinear programming: local convergence[J].SIAM J Optim, 2005, 16(1): 32-48.

[13] Nie Puyan. Sequential penalty quadratic programming filter methods for nonlinear programming[J]. Nonlinear Anal-Real, 2007, 8(1): 118-129.

[14] Li Duoquan. A new sqp-filter method for solving nonlinear programming problems[J]. Journal ofcomputational mathematics, 2006, 24(5): 609-634.

[15] 简金宝,光滑约束优化快速算法——理论分析与数值实验[M]. 北京, 科学出版社,2010.

[16] Ulbrich, Ulbrich M. Nonmonotone trust region methods for nonlinear equality constrained optimizationwithout a penalty function[J]. Mathematical Programming, Series B, 95(1), 103-105,2003.

[17] Gould N, Toint Ph L. Nonlinear programming without a penalty function or a filter[J]. MathProgram, 2010, 122(1): 155-196.

[18] Gould N, Robinson D P, Toint Ph L. Corrigendum: Nonlinear programming without a penaltyfunction or a filter, Technical Report, Namur Center for Complex Systems, 2011.

[19] Liu Xinwei, Yuan Yaxiang. A sequential quadratic programming method without a penalty functionor a filter for nonlinear equality constrained optimization[J]. SIAM J Optim, 2011, 21(2):545-571.

[20] Solodov M V. Global convergence of an SQP method without boundedness assumptions on anyof iterative sequences[J]. Math Program, 2009, 118: 1-12.

[21] Zhou Guanglu. A modified SQP method and its global convergence[J]. Journal of Global Optimization,1997, 11(2): 193-205.

[22] Hock W, Schittkowski K. Test examples for nonlinear Programming Codes. Lecture Notes inEconomics and Mathematical Systems[J]. Springer, 1981, 187.
[1] 郦旭东. 复合凸优化的快速邻近点算法[J]. 计算数学, 2020, 42(4): 385-404.
[2] 尚在久, 宋丽娜. 关于辛算法稳定性的若干注记[J]. 计算数学, 2020, 42(4): 405-418.
[3] 陈园. 无单调性集值变分不等式的一种投影算法[J]. 计算数学, 2020, 42(4): 435-444.
[4] 尹江华, 简金宝, 江羡珍. 凸约束非光滑方程组一个新的谱梯度投影算法[J]. 计算数学, 2020, 42(4): 457-471.
[5] 崔俊芝, 余翌帆. 基于分子动力学模拟的金属构件的弹-塑性分解方法[J]. 计算数学, 2020, 42(3): 279-297.
[6] 钟巍, 田宙, 寿列枫. 基于线性代数的大规模快速量纲分析算法及其在爆炸与冲击工程研究中的应用[J]. 计算数学, 2020, 42(2): 170-195.
[7] 张纯, 贾泽慧, 蔡邢菊. 广义鞍点问题的改进的类SOR算法[J]. 计算数学, 2020, 42(1): 39-50.
[8] 李枝枝, 柯艺芬, 储日升, 张怀. 二阶锥线性互补问题的广义模系矩阵分裂迭代算法[J]. 计算数学, 2019, 41(4): 395-405.
[9] 刘子源, 梁家瑞, 钱旭, 宋松和. 带乘性噪声的空间分数阶随机非线性Schrödinger方程的广义多辛算法[J]. 计算数学, 2019, 41(4): 440-452.
[10] 印卧涛. 无中心优化的算子分裂方法[J]. 计算数学, 2019, 41(3): 225-241.
[11] 申子慧, 申培萍. 线性分式多乘积规划问题的完全多项式时间近似算法[J]. 计算数学, 2019, 41(2): 212-218.
[12] 夏雨晴, 张振跃. 子空间聚类的重建模型及其快速算法[J]. 计算数学, 2019, 41(1): 1-11.
[13] 闫熙, 马昌凤. 求解矩阵方程AXB+CXD=F参数迭代法的最优参数分析[J]. 计算数学, 2019, 41(1): 37-51.
[14] 郭科, 韩德仁. 单调算子理论与分裂算法[J]. 计算数学, 2018, 40(4): 418-435.
[15] 蒋建林, 潘蕴文. 大规模多设施Weber问题的改进Cooper算法[J]. 计算数学, 2018, 40(4): 470-484.
阅读次数
全文


摘要