A Continuous Dynamic Stochastic Approximation Procedure | ||
The Egyptian Statistical Journal | ||
Article 10, Volume 37, Issue 1, June 1993, Pages 155-168 PDF (4.85 M) | ||
Document Type: Original Article | ||
DOI: 10.21608/esju.1993.426939 | ||
Author | ||
El Sayed Sorour* | ||
Dep. of Mathematics, Military Technical College, Kobry El-Kobba, Cairo, Egypt | ||
Abstract | ||
This paper considers the continuous Kiefer-Nolfowitz stochastic approximation procedure, where the regression function changes with time t. Let ED(t) be the unique minimum (maximum) of the regression function at a time t. Our goal is to select 0(t) such that ilx(t) - 0,t-0. co. It is assumed that 6)(t)= (91(t),...,913(0) and ei(t)=(B3,U(t)), j=1,...,P for B.(unknown) and U(t) (known) are elements of a Hilbert space H. Under general conditions we prove that lim flBi(t) - B.II<oo. If we restrict H to Rk and more t-)•co stringent conditions we prove that 11 x(t)- e(t)1k0,w.p.1 as | ||
Statistics Article View: 51 PDF Download: 17 |