Acceleration of stochastic approximation by averaging
SIAM Journal on Control and Optimization
A scaled stochastic approximation algorithm
Management Science
Retrospective simulation response optimization
WSC '91 Proceedings of the 23rd conference on Winter simulation
A projected stochastic approximation algorithm
WSC '91 Proceedings of the 23rd conference on Winter simulation
Feature Article: Optimization for simulation: Theory vs. Practice
INFORMS Journal on Computing
Introduction to Stochastic Search and Optimization
Introduction to Stochastic Search and Optimization
Variable-sample methods for stochastic optimization
ACM Transactions on Modeling and Computer Simulation (TOMACS)
Efficient sample sizes in stochastic nonlinear programming
Journal of Computational and Applied Mathematics
Retrospective-approximation algorithms for the multidimensional stochastic root-finding problem
ACM Transactions on Modeling and Computer Simulation (TOMACS)
Discrete stochastic optimization using linear interpolation
Proceedings of the 40th Conference on Winter Simulation
The stochastic root-finding problem: Overview, solutions, and open questions
ACM Transactions on Modeling and Computer Simulation (TOMACS)
An adaptive multidimensional version of the Kiefer-Wolfowitz stochastic approximation algorithm
Winter Simulation Conference
A Sequential Sampling Procedure for Stochastic Programming
Operations Research
Proceedings of the Winter Simulation Conference
Hi-index | 0.00 |
Retrospective Approximation (RA) is a solution paradigm introduced in the early 1990s by Chen and Schmeiser for solving one-dimensional stochastic root finding problems (SRFPs). The RA paradigm can be thought of as a refined and implementable version of sample average approximation, where a sequence of approximate problems are strategically generated and solved to identify iterates that progressively approach the desired solution. While originally aimed at one-dimensional SRFPs, the paradigm's broader utility, particularly within general simulation optimization algorithms, is becoming increasingly evident. We discuss the RA paradigm, demonstrate its usefulness, present the key results and papers on the topic over the last fifteen years, and speculate fruitful future directions.