Distributionally Constrained Black-Box Stochastic Gradient Estimation and Optimization

被引:0
作者
Lam, Henry [1 ]
Zhang, Junhui [2 ]
机构
[1] Columbia Univ, Dept Ind Engn & Operat Res, New York, NY 10027 USA
[2] Columbia Univ, Dept Appl Phys & Appl Math, New York, NY 10027 USA
基金
美国国家科学基金会;
关键词
zeroth-order gradient estimation; finite difference; simultaneous perturbation; distributionally robust optimization; stochastic approximation; ROBUST; SIMULATION; SENSITIVITY; RISK;
D O I
10.1287/opre.2021.0307
中图分类号
C93 [管理学];
学科分类号
12 ; 1201 ; 1202 ; 120202 ;
摘要
We consider stochastic gradient estimation using only black-box function evaluations, where the function argument lies within a probability simplex. This problem is motivated from gradient-descent optimization procedures in multiple applications in distributionally robust analysis and inverse model calibration involving decision variables that are probability distributions. We are especially interested in obtaining gradient estimators where one or few sample observations or simulation runs apply simultaneously to all directions. Conventional zeroth-order gradient schemes such as simultaneous perturbation face challenges as the required moment conditions that allow the "canceling" of higher- order biases cannot be satisfied without violating the simplex constraints. We investigate a new set of required conditions on the random perturbation generator, which leads us to a class of implementable gradient estimators using Dirichlet mixtures. We study the statistical properties of these estimators and their utility in constrained stochastic approximation. We demonstrate the effectiveness of our procedures and compare with benchmarks via several numerical examples.
引用
收藏
页数:16
相关论文
共 53 条
[1]  
[Anonymous], 2013, Monte Carlo methods in financial engineering
[2]  
Asmussen S., 2007, STOCHASTIC SIMULATIO
[3]   Model calibration via distributionally robust optimization: On the NASA Langley Uncertainty Quantification Challenge [J].
Bai, Yuanlu ;
Huang, Zhiyuan ;
Lam, Henry .
MECHANICAL SYSTEMS AND SIGNAL PROCESSING, 2022, 164
[4]   Resampling methods for input modeling [J].
Barton, RR ;
Schruben, LW .
WSC'01: PROCEEDINGS OF THE 2001 WINTER SIMULATION CONFERENCE, VOLS 1 AND 2, 2001, :372-378
[5]  
Barton RR, 2022, The Palgrave Handbook of Operations Research, P573
[6]   Mirror descent and nonlinear projected subgradient methods for convex optimization [J].
Beck, A ;
Teboulle, M .
OPERATIONS RESEARCH LETTERS, 2003, 31 (03) :167-175
[7]   Robust Solutions of Optimization Problems Affected by Uncertain Probabilities [J].
Ben-Tal, Aharon ;
den Hertog, Dick ;
De Waegenaere, Anja ;
Melenberg, Bertrand ;
Rennen, Gijs .
MANAGEMENT SCIENCE, 2013, 59 (02) :341-357
[8]  
BenTal A, 2009, PRINC SER APPL MATH, P1
[9]   A Theoretical and Empirical Comparison of Gradient Approximations in Derivative-Free Optimization [J].
Berahas, Albert S. ;
Cao, Liyuan ;
Choromanski, Krzysztof ;
Scheinberg, Katya .
FOUNDATIONS OF COMPUTATIONAL MATHEMATICS, 2022, 22 (02) :507-560
[10]   Robust sample average approximation [J].
Bertsimas, Dimitris ;
Gupta, Vishal ;
Kallus, Nathan .
MATHEMATICAL PROGRAMMING, 2018, 171 (1-2) :217-282