Using perturbation analysis for gradient estimation, averaging and updating in a stochastic approximation algorithm

Authors:
Michael C. Fu;Yu-Chi Ho
Affiliations:
Division of Applied Sciences, Harvard University, Cambridge, MA;Division of Applied Sciences, Harvard University, Cambridge, MA
Venue:
WSC '88 Proceedings of the 20th conference on Winter simulation
Year:
1988

Citing 5
Cited 6

Infinitesimal perturbation analysis for general discrete event systems

Journal of the ACM (JACM)
A sample performance function of closed Jackson queueing networks

Operations Research
Perturbation analysis gives strongly consistent sensitivity estimates for the M/G/1 queue

Management Science
Optimization of stochastic systems

WSC '86 Proceedings of the 18th conference on Winter simulation
Optimization in simulation: a survey of recent results

WSC '87 Proceedings of the 19th conference on Winter simulation

Simulation optimization of (s,S) inventory systems

WSC '92 Proceedings of the 24th conference on Winter simulation
Simulation optimization via simultaneous perturbation stochastic approximation

WSC '94 Proceedings of the 26th conference on Winter simulation
Transfer optimization via simultaneous perturbation stochastic approximation

WSC '95 Proceedings of the 27th conference on Winter simulation
Simulation optimization: a survey of simulation optimization techniques and procedures

Proceedings of the 32nd conference on Winter simulation
Central Limit Theorems for Stochastic Optimization AlgorithmsUsing Infinitesimal Perturbation Analysis

Discrete Event Dynamic Systems
Applying perturbation analysis to traffic shaping

Computer Communications

Quantified Score

Hi-index	0.00

Visualization

Abstract

We propose a gradient updating procedure for using both “present” and “past” data to improve the convergence properties of a stochastic approximation algorithm. This procedure utilizes second derivatives estimated by perturbation analysis techniques. Experimental evidence provided by simulation runs appear to confirm the improvement in convergence rate gained by this modified algorithm.