Quasi-Monte Carlo strategies for stochastic optimization

Authors:
Shane S. Drew;Tito Homem-de-Mello
Affiliations:
Northwestern University, Evanston, IL;Northwestern University, Evanston, IL
Venue:
Proceedings of the 38th conference on Winter simulation
Year:
2006

Citing 6
Cited 1

Random number generation and quasi-Monte Carlo methods

Random number generation and quasi-Monte Carlo methods
Monte Carlo (importance) sampling within a Benders decomposition algorithm for stochastic linear programs

Annals of Operations Research - Special issue on sensitivity analysis and optimization of discrete event systems
An efficient sampling technique for off-line quality control

Technometrics
Latin supercube sampling for very high-dimensional simulations

ACM Transactions on Modeling and Computer Simulation (TOMACS) - Special issue on uniform random number generation
Decomposition Algorithms for Stochastic Programming on a Computational Grid

Computational Optimization and Applications
Variance reduction in sample approximations of stochastic programs

Mathematical Programming: Series A and B

Random search for hyper-parameter optimization

The Journal of Machine Learning Research

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper we discuss the issue of solving stochastic optimization problems using sampling methods. Numerical results have shown that using variance reduction techniques from statistics can result in significant improvements over Monte Carlo sampling in terms of the number of samples needed for convergence of the optimal objective value and optimal solution to a stochastic optimization problem. Among these techniques are stratified sampling and Quasi-Monte Carlo sampling. However, for problems in high dimension, it may be computationally inefficient to calculate Quasi-Monte Carlo point sets in the full dimension. Rather, we wish to identify which dimensions are most important to the convergence and implement a Quasi-Monte Carlo sampling scheme with padding, where the important dimensions are sampled via Quasi-Monte Carlo sampling and the remaining dimensions with Monte Carlo sampling. We then incorporate this sampling scheme into an external sampling algorithm (ES-QMCP) to solve stochastic optimization problems.