An ISS and l-stability approach to forward error analysis of iterative numerical algorithms

In this paper we have used results from ISS and lοο stability theory to find a systematic way of finding bounds on the forward error and the maximum finite precision error. The proposed approach is applied on the successive iteration method and the classical iterative methods. The obtained forward e...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:49th IEEE Conference on Decision and Control (CDC) S. 780 - 785
Hauptverfasser: Hasan, A, Kerrigan, E C, Constantinides, G A
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 01.12.2010
Schlagworte:
Algorithm design and analysis
Approximation algorithms
Equations
Heuristic algorithms
Iterative methods
Jacobian matrices
Mathematical model
ISBN:142447745X, 9781424477456
ISSN:0191-2216
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:In this paper we have used results from ISS and lοο stability theory to find a systematic way of finding bounds on the forward error and the maximum finite precision error. The proposed approach is applied on the successive iteration method and the classical iterative methods. The obtained forward error bound is numerically observed to be tighter than the bounds given in the numerical analysis literature. Moreover, in contrast to bounds in the numerical analysis literature for the successive iteration method, the bound obtained using a control approach does not involve the solution of the problem. Although the proposed scheme provides a new systematic way for forward error analysis, we note that the conditions of ISS and lοο-stability might be restrictive and not all algorithms will satisfy them.
ISBN:142447745X
9781424477456
ISSN:0191-2216
DOI:10.1109/CDC.2010.5717884