Randomization methods in optimization and adaptive control - Dedicated to Tyrone Duncan on occasion of his 60th birthday
2002 (English)In: STOCHASTIC THEORY AND CONTROL: Proceedings of a Workshop held in Lawrence, Kansas, Springer-Verlag New York, 2002, Vol. 280, 137-153 p.Chapter in book (Refereed)
We consider simultaneous perturbation stochastic approximation (SPSA) methods applied to noise-free problems in optimization and adaptive control. More generally, we consider discrete-time fixed gain stochastic approximation processes that are defined in terms of a random field that is identically zero at some point theta*. The boundedness of the estimator process is enforced by a resetting mechanism. Under appropriate technical conditions the estimator sequence converges to theta* with geometric rate almost surely. This result is in striking contrast to classical stochastic approximation theory where the typical convergence rate is n(-1/2). For the proof a discrete-time version of the ODE-method is used and the techniques of  are extended. A simple variant of noise free-SPSA is applied to extend a direct controller tuning method named Iterative Feedback Timing (IFT), see . Using randomization, the number of experiments required to obtain an unbiased estimate of the gradient of the cost function can be reduced significantly for multi-input multi-output systems.
Place, publisher, year, edition, pages
Springer-Verlag New York, 2002. Vol. 280, 137-153 p.
, LECTURE NOTES IN CONTROL AND INFORMATION SCIENCES, ISSN 0170-8643
PERTURBATION GRADIENT APPROXIMATION, STOCHASTIC-APPROXIMATION, CONVERGENCE, ALGORITHM
Research subject SRA - ICT
IdentifiersURN: urn:nbn:se:kth:diva-26598ISI: 000177472100011ISBN: 978-3-540-43777-0OAI: oai:DiVA.org:kth-26598DiVA: diva2:376353
Workshop on Stochastic Theory and Control UNIV KANSAS, LAWRENCE, KS, OCT 18-20, 2001
QC 20101210 NR 201408042010-12-102010-11-252012-01-13Bibliographically approved