• 论文 • 上一篇    下一篇

二次插值模型直接搜索算法的参数分析

刘浩1, 倪勤2   

    1. 南京航空航天大学经济管理学院, 南京 210016;  南京工业大学 理学院, 南京 210009
    2. 南京航空航天大学理学院, 南京 210016
  • 出版日期:2008-12-14 发布日期:2008-12-24
  • 基金资助:

    国家自然科学基金(10471062);  江苏省基础研究计划(自然科学基金 BK2006184)资助项目.

刘浩, 倪勤. 二次插值模型直接搜索算法的参数分析[J]. 数值计算与计算机应用, 2008, 29(4): 266-276.

Liu Hao, Ni Qin. SENSITIVITY  OF PARAMETERS IN DIRECT SEARCH METHODS BASED ON  QUADRATIC INTERPOLATION MODEL[J]. Journal of Numerical Methods and Computer Applications, 2008, 29(4): 266-276.

SENSITIVITY  OF PARAMETERS IN DIRECT SEARCH METHODS BASED ON  QUADRATIC INTERPOLATION MODEL

Liu Hao1, Ni Qin2   

    1. College of Economics & Management Science, Nanjing University of  Aeronautics and Astronautics, Nanjing 210016, China, ;
      College of Science, Nanjing University of Technology, Nanjing 210009,  China
    2. College of Science, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China
  • Online:2008-12-14 Published:2008-12-24

对无约束优化问题的二次插值型直接搜索算法中初始插值半径,信赖域初始半径, 位移接受准则和信赖域半径调节参数进行了数值实验分析.
数值实验表明解无约束优化的基于二次函数插值型的直接搜索算法对初始插值半径和信赖域初始半径比较敏感,对位移接受准则和半径调节参数不敏感. 根据数值实验结果推荐初始插值半径的选取应与信赖域初始半径相等,同时 给出了基于二次插值型的直接搜索算法中初始插值半径与信赖域初始半径的选择区间和其它参数的推荐值. 这些结果对这类算法的数值实现和工程应用是有益的.

The sensitivity of parameters  in direct search algorithm with quadratic interpolation model is examined by numerical experiments. These
parameters are related to the initial radius, the step acceptance and the update of the trust region. According to numerical experiments, it is shown that the numerical efficiency of the direct search algorithm is very sensitive to the initial radius, and is insensitive to the parameters related to the step acceptance and the update of the trust region. Numerical tests show that initial interpolation radius should be equal to initial trust region radius. Recommended ranges of values for initial radius and the values of other parameters are exhibited on the basis of extensive numerical
tests, which is beneficial to the implementation of the algorithm and engineering application.

MR(2010)主题分类: 

()

[1]Levenberg K. A method for the solution of certain problms in least square. Quarterly Journal on Applied Mathematics, 1944, 2: 164-168.
[2]More J J. Recent developments in algorithms and software for trust region methods. In: A. Bachem, M. Grotschel and B. Korte (eds), Mathematical Programming: The state of the art. Berlin Heidelberg New York, Springer, 1983, 258-287.
[3]袁亚湘,孙文瑜. 最优化理论与方法[M]. 北京: 科学出版社, 1997.
[4]Powell M J D. A direct search optimization method that models the objective and constrained functions by linear interpolation. In: S. Gomez and J. P. Hennart(eds), Advances in Optimization and Numerical Analysis. Dordrecht, Kluwer Academic, 1994, 51-67.
[5]Powell M J D. Direct search algorithms for optimization calculations[J]. Acta Numerica, 1998, 7: 287-336.
[6]Conn A R and Toint Ph L. An algorithm using quadratic interpolation for unconstrained optimization. In: G.Di Pillo, F. Gianessi (eds), Nonlinear Optimization and Applications. New York, Plenum Publishing, 1996, 27-47.
[7]Conn A R, Scheinberg K and Toint Ph L. On the convergence of derivative free methods for unconstrained optimization. In: A. Iserles and M. Buhmann (eds), Approximation Theory and Optimization: Tributes to M.J.D. Powell. Cambridge, Cambridge University Press, 1997, 83-108.
[8]Conn A R, Scheinberg K and Toint Ph L. Recent progress in unconstrained nonlinear optimization without derivatives, Math. Program., 1997, 97: 397-414.
[9] Ni Q and Hu S H. A new derivative free optimization method based on conic interpolation model[J]. ACTA Mathematica Scientia, 2004, 24: 281-290.
[10] Powell M J D. UOBYQA: unconstrained optimization by quadratic approximation[J]. Math. Program., 2002, 92: 555-582.
[11]Gould N I M, Orban D, Sartenaer A and Toint Ph L. Sensitivity of trust-region algorithms to their parameters[J]. 4OR, 2005, 3: 227-241.
[12]Sartenaer A. Automatic determination of an initial trust region in nonlinear programming[J]. SIAM J. Sci. Comput., 1997, 18: 1788-1803.
[13]Sauer Th and Yuan X. On multivariate polynimial interpolation[J]. Mathematics of Computation, 1995, 64: 1147-1170.
[14]More J J and Sorense D C. Computing a trust region step[J]. SIAM J. Sci. Stat. Comput., 1983, 4: 553-572.
[15]孙文瑜,徐成贤,朱德通. 最优化方法[M]. 北京: 高等教育出版社, 2003. >br>[16]Luksan L and Vlcek J. Test problems for unconstrained optimization[R]. Technique Report 2003 No.897, of Insitiute of Computer Science, Academy of Sciences of the Czech Republic.
[17]Conn A R, Gould N I M and Toint Ph L. LANCELOT: a Fortran package for large-scale nonlinear optimization( Release A). Springer Series in Computational Mathematics, No. 17, Springer, Berlin Heidelberg New York, 1992.
[1] 王丽平, 吴亚飞. 联合稀疏独立法则及其在疾病分类中的应用[J]. 数值计算与计算机应用, 2016, 37(3): 179-185.
[2] 张艳君, 赵金玲, 徐尔. 求解多集分裂可行问题的一种共轭梯度法[J]. 数值计算与计算机应用, 2013, 34(4): 249-256.
[3] 陈龙卫, 倪勤, 张欣. 强迫下降的三项共扼梯度法[J]. 数值计算与计算机应用, 2012, 33(3): 181-188.
[4] 徐海文. 一类单调非线性变分不等式的前向加速收缩算法[J]. 数值计算与计算机应用, 2011, 32(4): 259-266.
[5] 马明, 刘华. 基于最小平方误差逼近的线性阀体的设计[J]. 数值计算与计算机应用, 2010, 31(3): 214-222.
[6] 胡平, 倪勤. 一个松弛非单调线搜索法[J]. 数值计算与计算机应用, 2010, 31(3): 191-199.
[7] 刘金魁, 王开荣, 杜祥林, 贾松芳. 一种新的非线性共轭梯度法及收敛性[J]. 数值计算与计算机应用, 2009, 30(4): 247-254.
[8] 孙莉, 贺国平, 房亮. 基于求解大规模界约束问题的三种有效集识别策略的比较[J]. 数值计算与计算机应用, 2009, 30(1): 41-47.
阅读次数
全文


摘要