ON THE WORST-CASE DIVERGENCE OF THE LEAST-SQUARES ALGORITHM

Authors
AKCAY H NINNESS B
Citation
H. Akcay et B. Ninness, ON THE WORST-CASE DIVERGENCE OF THE LEAST-SQUARES ALGORITHM, Systems & control letters, 33(1), 1998, pp. 19-24
Citations number
23
Categorie Soggetti
Robotics & Automatic Control","Operatione Research & Management Science","Robotics & Automatic Control","Operatione Research & Management Science
Journal title
Systems & control letters → ACNP
ISSN journal
01676911
Volume
33
Issue
1
Year of publication
1998
Pages
19 - 24
Database
ISI
SICI code
0167-6911(1998)33:1<19:OTWDOT>2.0.ZU;2-3
Abstract
In this paper, we provide a H-infinity-norm lower bound on the worst-c ase identification error of least-squares estimation when using FIR mo del structures. This bound increases as a logarithmic function of mode l complexity and is valid for a wide class of inputs characterized as being quasi-stationary with covariance function falling off sufficient ly quickly. (C) 1998 Elsevier Science B.V. All rights reserved.