A-OPTIMAL DESIGN OF EXPERIMENTS FOR INFINITE-DIMENSIONAL BAYESIAN LINEAR INVERSE PROBLEMS WITH REGULARIZED l0-SPARSIFICATION

被引:87
作者
Alexanderian, Alen [1 ]
Petra, Noemi [1 ]
Stadler, Georg [1 ]
Ghattas, Omar [2 ]
机构
[1] Univ Texas Austin, Inst Computat Engn & Sci, Austin, TX 78712 USA
[2] Univ Texas Austin, Inst Computat Engn & Sci, Dept Geol Sci, Dept Mech Engn, Austin, TX 78712 USA
基金
美国国家科学基金会;
关键词
optimal experimental design; A-optimal design; Bayesian inference; sensor placement; ill-posed inverse problems; low-rank approximation; randomized trace estimator; randomized SVD; ALGORITHMS; MATRIX;
D O I
10.1137/130933381
中图分类号
O29 [应用数学];
学科分类号
070104 ;
摘要
We present an efficient method for computing A-optimal experimental designs for infinite-dimensional Bayesian linear inverse problems governed by partial differential equations (PDEs). Specifically, we address the problem of optimizing the location of sensors (at which observational data are collected) to minimize the uncertainty in the parameters estimated by solving the inverse problem, where the uncertainty is expressed by the trace of the posterior covariance. Computing optimal experimental designs (OEDs) is particularly challenging for inverse problems governed by computationally expensive PDE models with infinite-dimensional (or, after discretization, high-dimensional) parameters. To alleviate the computational cost, we exploit the problem structure and build a low-rank approximation of the parameter-to-observable map, preconditioned with the square root of the prior covariance operator. The availability of this low-rank surrogate, relieves our method from expensive PDE solves when evaluating the optimal experimental design objective function, i.e., the trace of the posterior covariance, and its derivatives. Moreover, we employ a randomized trace estimator for efficient evaluation of the OED objective function. We control the sparsity of the sensor configuration by employing a sequence of penalty functions that successively approximate the l(0)-"norm"; this results in binary designs that characterize optimal sensor locations. We present numerical results for inference of the initial condition from spatiotemporal observations in a time-dependent advection-diffusion problem in two and three space dimensions. We find that an optimal design can be computed at a cost, measured in number of forward PDE solves, that is independent of the parameter and sensor dimensions. Moreover, the numerical optimization problem for finding the optimal design can be solved in a number of interior-point quasi-Newton iterations that is insensitive to the parameter and sensor dimensions. We demonstrate numerically that l(0)-sparsified experimental designs obtained via a continuation method outperform l(1)-sparsified designs.
引用
收藏
页码:A2122 / A2148
页数:27
相关论文
共 34 条
  • [11] COMPUTING f(A)b VIA LEAST SQUARES POLYNOMIAL APPROXIMATIONS
    Chen, Jie
    Anitescu, Mihai
    Saad, Yousef
    [J]. SIAM JOURNAL ON SCIENTIFIC COMPUTING, 2011, 33 (01) : 195 - 222
  • [12] Deuflhard P., 2004, SPRINGER SER COMPUT, V35
  • [13] Compressed sensing
    Donoho, DL
    [J]. IEEE TRANSACTIONS ON INFORMATION THEORY, 2006, 52 (04) : 1289 - 1306
  • [14] FAST ALGORITHMS FOR BAYESIAN UNCERTAINTY QUANTIFICATION IN LARGE-SCALE LINEAR INVERSE PROBLEMS BASED ON LOW-RANK PARTIAL HESSIAN APPROXIMATIONS
    Flath, H. P.
    Wilcox, L. C.
    Akcelik, V.
    Hill, J.
    Waanders, B. van Bloemen
    Ghattas, O.
    [J]. SIAM JOURNAL ON SCIENTIFIC COMPUTING, 2011, 33 (01) : 407 - 432
  • [15] Golub G., 2013, J HOPKINS STUDIES MA
  • [16] Numerical methods for the design of large-scale nonlinear discrete ill-posed inverse problems
    Haber, E.
    Horesh, L.
    Tenorio, L.
    [J]. INVERSE PROBLEMS, 2010, 26 (02)
  • [17] Numerical methods for experimental design of large-scale linear ill-posed inverse problems
    Haber, E.
    Horesh, L.
    Tenorio, L.
    [J]. INVERSE PROBLEMS, 2008, 24 (05)
  • [18] Haber E., 2012, COMPUT OPTIM APPL, P1
  • [19] Finding Structure with Randomness: Probabilistic Algorithms for Constructing Approximate Matrix Decompositions
    Halko, N.
    Martinsson, P. G.
    Tropp, J. A.
    [J]. SIAM REVIEW, 2011, 53 (02) : 217 - 288
  • [20] Horesh L, 2011, WILEY SERIES COMPUTA, P273