Optimization via simulation: randomized-direction stochastic approximation algorithms using deterministic sequences

Authors:
Xiaoping Xiong;I-Jeng Wang;Michael C. Fu
Affiliations:
University of Maryland, College Park, MD;Applied Physics Laboratory, Laurel, MD;University of Maryland, College Park, MD
Venue:
Proceedings of the 34th conference on Winter simulation: exploring new frontiers
Year:
2002

Citing 4
Cited 3

Experiments in nonconvex optimization: stochastic approximation with function smoothing and simulated annealing

Neural Networks
Random number generation and quasi-Monte Carlo methods

Random number generation and quasi-Monte Carlo methods
A one-measurement form of simultaneous perturbation stochastic approximation

Automatica (Journal of IFAC)
Two-timescale simultaneous perturbation stochastic approximation using deterministic perturbation sequences

ACM Transactions on Modeling and Computer Simulation (TOMACS)

Universal parameter optimisation in games based on SPSA

Machine Learning
SPSA algorithms with measurement reuse

Proceedings of the 38th conference on Winter simulation
Stochastic gradient estimation using a single design point

Proceedings of the 38th conference on Winter simulation

Quantified Score

Hi-index	0.00

Visualization

Abstract

We study the convergence and asymptotic normality of a generalized form of stochastic approximation algorithm with deterministic perturbation sequences. Both one-simulation and two-simulation methods are considered. Assuming a special structure of deterministic sequence, we establish sufficient condition on the noise sequence for a.s. convergence of the algorithm. Construction of such a special structure of deterministic sequence follows the discussion of asymptotic normality. Finally we discuss ideas on further research in analysis and design of the deterministic perturbation sequences.