AN ALGORITHM FOR STOCHASTIC-APPROXIMATION WITH A PREASSIGNED PROBABILITY OF NOT EXCEEDING A REQUIRED THRESHOLD

Authors
GUSEV SV KRASULINA TP
Citation
Sv. Gusev et Tp. Krasulina, AN ALGORITHM FOR STOCHASTIC-APPROXIMATION WITH A PREASSIGNED PROBABILITY OF NOT EXCEEDING A REQUIRED THRESHOLD, Journal of computer & systems sciences international, 33(5), 1995, pp. 39-41
Citations number
7
Categorie Soggetti
Controlo Theory & Cybernetics","Computer Science Theory & Methods
Journal title
Journal of computer & systems sciences international → ACNP
ISSN journal
10642307
Volume
33
Issue
5
Year of publication
1995
Pages
39 - 41
Database
ISI
SICI code
1064-2307(1995)33:5<39:AAFSWA>2.0.ZU;2-C
Abstract
The one-sided convergence of the modified Robbins-Monro procedure intr oduced by Anbar is investigated. A method of choosing the constants in the algorithm that ensures a probability of not exceeding the require d root of the equation of regression as close to unity as desired is p roposed.