Sufficient conditions for the convergence of nonautonomous stochastic search for a global minimum

Bibliographic description

title:	Sufficient conditions for the convergence of nonautonomous stochastic search for a global minimum
author:	Tarłowski Dawid
journal title:	Universitatis Iagellonicae Acta Mathematica
volume:	49
date of publication :	2011
pages:	73-83
ISSN:	0083-4386
eISSN:	2084-3828
DOI:	10.4467/20843828AM.12.005.0457
language:	English
journal language:	English
abstract in English:	The majority of stochastic optimization algorithms can be writ- ten in the general form $x_{t+1}= T_{t}\left(x_{t},y_{t}\right)$, where $x_{t}$ is a sequence of points and parameters which are transformed by the algorithm, $T^{_{t}}$ are the methods of the algorithm and $Y_{t}$ represent the randomness of the algorithm. We extend the results of papers [11] and [14] to provide some new general conditions under which the algorithm finds a global minimum with probability one.
keywords in English:	stochastic optimization, global optimization, Lyapunov function, weak convergence of measures
affiliation:	Wydział Matematyki i Informatyki : Instytut Matematyki
type:	journal article
subtype:	academic paper