INEXACT OBJECTIVE FUNCTION EVALUATIONS IN A TRUST-REGION ALGORITHM FOR PDE-CONSTRAINED OPTIMIZATION UNDER UNCERTAINTY

被引:57
作者
Kouri, D. P. [1 ]
Heinkenschloss, M. [2 ]
Ridzal, D. [1 ]
Waanders, B. G. Van Bloemen [1 ]
机构
[1] Sandia Natl Labs, Albuquerque, NM 87185 USA
[2] Rice Univ, Dept Computat & Appl Math, Houston, TX 77005 USA
基金
美国国家科学基金会;
关键词
PDE-constrained optimization; uncertainty; stochastic collocation; trust regions; sparse grids; adaptivity; DIFFERENTIAL-EQUATIONS; STOCHASTIC COLLOCATION; INTEGRATION;
D O I
10.1137/140955665
中图分类号
O29 [应用数学];
学科分类号
070104 ;
摘要
This paper improves the trust-region algorithm with adaptive sparse grids introduced in [SIAM J. Sci. Comput., 35 (2013), pp. A1847-A1879] for the solution of optimization problems governed by partial differential equations (PDEs) with uncertain coefficients. The previous algorithm used adaptive sparse-grid discretizations to generate models that are applied in a trust-region framework to generate a trial step. The decision whether to accept this trial step as the new iterate, however, required relatively high-fidelity adaptive discretizations of the objective function. In this paper, we extend the algorithm and convergence theory to allow the use of low-fidelity adaptive sparse-grid models in objective function evaluations. This is accomplished by extending conditions on inexact function evaluations used in previous trust-region frameworks. Our algorithm adaptively builds two separate sparse grids: one to generate optimization models for the step computation and one to approximate the objective function. These adapted sparse grids often contain significantly fewer points than the high-fidelity grids, which leads to a dramatic reduction in the computational cost. This is demonstrated numerically using two examples. Moreover, the numerical results indicate that the new algorithm rapidly identifies the stochastic variables that are relevant to obtaining an accurate optimal solution. When the number of such variables is independent of the dimension of the stochastic space, the algorithm exhibits near dimension-independent behavior.
引用
收藏
页码:A3011 / A3029
页数:19
相关论文
共 28 条
[1]   A trust-region framework for managing the use of approximation models in optimization [J].
Alexandrov, NM ;
Dennis, JE ;
Lewis, RM ;
Torczon, V .
STRUCTURAL OPTIMIZATION, 1998, 15 (01) :16-23
[2]  
[Anonymous], TRUST REGION METHODS, DOI DOI 10.1137/1.9780898719857
[3]  
[Anonymous], 1999, INTERPOLATION SPATIA
[4]   A Stochastic Collocation Method for Elliptic Partial Differential Equations with Random Input Data [J].
Babuska, Ivo ;
Nobile, Fabio ;
Tempone, Raul .
SIAM REVIEW, 2010, 52 (02) :317-355
[5]   High dimensional polynomial interpolation on sparse grids [J].
Barthelmann, V ;
Novak, E ;
Ritter, K .
ADVANCES IN COMPUTATIONAL MATHEMATICS, 2000, 12 (04) :273-288
[6]   A retrospective trust-region method for unconstrained optimization [J].
Bastin, Fabian ;
Malmedy, Vincent ;
Mouffe, Melodie ;
Toint, Philippe L. ;
Tomanos, Dimitri .
MATHEMATICAL PROGRAMMING, 2010, 123 (02) :395-418
[7]  
Bochev P, 2012, SCI PROGRAMMING-NETH, V20, P151, DOI [10.3233/SPR-2012-0340, 10.1155/2012/403902]
[8]  
Carter, 1989, 8945 ICASE
[9]   NUMERICAL EXPERIENCE WITH A CLASS OF ALGORITHMS FOR NONLINEAR OPTIMIZATION USING INEXACT FUNCTION AND GRADIENT INFORMATION [J].
CARTER, RG .
SIAM JOURNAL ON SCIENTIFIC COMPUTING, 1993, 14 (02) :368-388
[10]   ON THE GLOBAL CONVERGENCE OF TRUST REGION ALGORITHMS USING INEXACT GRADIENT INFORMATION [J].
CARTER, RG .
SIAM JOURNAL ON NUMERICAL ANALYSIS, 1991, 28 (01) :251-265