Efficient Iterated Filtering

Erik Lindström; Edward Ionides; Jan Frydendall; Henrik Madsen

doi:10.3182/20120711-3-BE-2027.00300

Efficient Iterated Filtering

Erik Lindström, Edward Ionides, Jan Frydendall, Henrik Madsen

CERE – Center for Energy Ressources Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference abstract in proceedings › Research › peer-review

Abstract

Parameter estimation in general state space models is not trivial as the likelihood is unknown. We propose a recursive estimator for general state space models, and show that the estimates converge to the true parameters with probability one. The estimates are also asymptotically Cramer-Rao efficient. The proposed estimator is easy to implement as it only relies on non-linear filtering. This makes the framework flexible as it is easy to tune the implementation to achieve computational efficiency. This is done by using the approximation of the score function derived from the theory on Iterative Filtering as a building block within the recursive maximum likelihood estimator.

Original language	English
Title of host publication	System Identification
Volume	16
Publisher	International Federation of Automatic Control
Publication date	2012
Pages	1785-1790
ISBN (Print)	978-3-902823-06-9
DOIs	https://doi.org/10.3182/20120711-3-BE-2027.00300
Publication status	Published - 2012
Event	16th IFAC Symposium on System Identification - Square - Brussels Meeting Centre, Brussels, Belgium Duration: 11 Jul 2012 → 13 Jul 2012 http://www.sysid2012.org/

Conference

Conference	16th IFAC Symposium on System Identification
Location	Square - Brussels Meeting Centre
Country/Territory	Belgium
City	Brussels
Period	11/07/2012 → 13/07/2012
Internet address	http://www.sysid2012.org/

Series	IFAC Proceedings Volumes (IFAC-PapersOnline)

Keywords

Approximation theory
Iterative methods
Maximum likelihood estimation
Parameter estimation
State space methods
Estimation

Access to Document

10.3182/20120711-3-BE-2027.00300

OpenUrl availability

Full text

Cite this

@inbook{b8aa341486754b5ab3df8f95122c5e3e,

title = "Efficient Iterated Filtering",

abstract = "Parameter estimation in general state space models is not trivial as the likelihood is unknown. We propose a recursive estimator for general state space models, and show that the estimates converge to the true parameters with probability one. The estimates are also asymptotically Cramer-Rao efficient. The proposed estimator is easy to implement as it only relies on non-linear filtering. This makes the framework flexible as it is easy to tune the implementation to achieve computational efficiency. This is done by using the approximation of the score function derived from the theory on Iterative Filtering as a building block within the recursive maximum likelihood estimator.",

keywords = "Approximation theory, Iterative methods, Maximum likelihood estimation, Parameter estimation, State space methods, Estimation",

author = "Erik Lindstr{\"o}m and Edward Ionides and Jan Frydendall and Henrik Madsen",

year = "2012",

doi = "10.3182/20120711-3-BE-2027.00300",

language = "English",

isbn = "978-3-902823-06-9",

volume = "16",

series = "IFAC Proceedings Volumes (IFAC-PapersOnline) ",

publisher = "International Federation of Automatic Control",

pages = "1785--1790",

booktitle = "System Identification",

note = "16th IFAC Symposium on System Identification, SYSID 2012 ; Conference date: 11-07-2012 Through 13-07-2012",

url = "http://www.sysid2012.org/",

}

TY - ABST

T1 - Efficient Iterated Filtering

AU - Lindström, Erik

AU - Ionides, Edward

AU - Frydendall, Jan

AU - Madsen, Henrik

PY - 2012

Y1 - 2012

N2 - Parameter estimation in general state space models is not trivial as the likelihood is unknown. We propose a recursive estimator for general state space models, and show that the estimates converge to the true parameters with probability one. The estimates are also asymptotically Cramer-Rao efficient. The proposed estimator is easy to implement as it only relies on non-linear filtering. This makes the framework flexible as it is easy to tune the implementation to achieve computational efficiency. This is done by using the approximation of the score function derived from the theory on Iterative Filtering as a building block within the recursive maximum likelihood estimator.

AB - Parameter estimation in general state space models is not trivial as the likelihood is unknown. We propose a recursive estimator for general state space models, and show that the estimates converge to the true parameters with probability one. The estimates are also asymptotically Cramer-Rao efficient. The proposed estimator is easy to implement as it only relies on non-linear filtering. This makes the framework flexible as it is easy to tune the implementation to achieve computational efficiency. This is done by using the approximation of the score function derived from the theory on Iterative Filtering as a building block within the recursive maximum likelihood estimator.

KW - Approximation theory

KW - Iterative methods

KW - Maximum likelihood estimation

KW - Parameter estimation

KW - State space methods

KW - Estimation

U2 - 10.3182/20120711-3-BE-2027.00300

DO - 10.3182/20120711-3-BE-2027.00300

M3 - Conference abstract in proceedings

SN - 978-3-902823-06-9

VL - 16

T3 - IFAC Proceedings Volumes (IFAC-PapersOnline)

SP - 1785

EP - 1790

BT - System Identification

PB - International Federation of Automatic Control

T2 - 16th IFAC Symposium on System Identification

Y2 - 11 July 2012 through 13 July 2012

ER -

Efficient Iterated Filtering

Abstract

Conference

Keywords

Access to Document

OpenUrl availability

Fingerprint

Cite this