A direct filter method for parameter estimation

Richard Archibald; Feng Bao; Xuemin Tu

doi:10.1016/j.jcp.2019.108871

A direct filter method for parameter estimation

Richard Archibald, Feng Bao, Xuemin Tu

Data Analysis and Machine Learning

Research output: Contribution to journal › Article › peer-review

11 Scopus citations

Abstract

Parameter estimation is an important research topic in data assimilation. In this paper, a novel parameter estimation method is introduced, where the parameter is considered as the state process in a nonlinear filtering problem and the state model that contains the parameter is used to construct a pseudo-observation. This approach is named the direct filter method since nonlinear filtering algorithms are used to estimate the parameter directly without estimating the state model as part of the solution in the nonlinear filtering problem. Numerical experiments are carried out to examine the effectiveness and accuracy of the direct filter method.

Original language	English
Article number	108871
Journal	Journal of Computational Physics
Volume	398
DOIs	https://doi.org/10.1016/j.jcp.2019.108871
State	Published - Dec 1 2019

Funding

This work is partially supported by the Scientific Discovery through Advanced Computing (SciDAC) program funded by U.S. Department of Energy, Office of Science, Advanced Scientific Computing Research through FASTMath Institute and CompFUSE project. The second author also acknowledges support by U.S. National Science Foundation under Contract DMS-1720222. The third author acknowledges support by U.S. National Science Foundation under Contracts DMS-1419069 and DMS-1723066. This work is partially supported by the Scientific Discovery through Advanced Computing (SciDAC) program funded by U.S. Department of Energy , Office of Science, Advanced Scientific Computing Research through FASTMath Institute and CompFUSE project. The second author also acknowledges support by U.S. National Science Foundation under Contract DMS-1720222 . The third author acknowledges support by U.S. National Science Foundation under Contracts DMS-1419069 and DMS-1723066 .

Funders	Funder number
FASTMath Institute
Scientific Discovery
U.S. National Science Foundation	DMS-1419069, DMS-1720222, DMS-1723066
U.S. Department of Energy
Office of Science
Advanced Scientific Computing Research

Keywords

Bayesian inference
Data assimilation
Nonlinear filtering problem
Parameter estimation
State-space model

Access to Document

10.1016/j.jcp.2019.108871

Cite this

@article{c4ba674fe8094135ab614b538750b222,

title = "A direct filter method for parameter estimation",

abstract = "Parameter estimation is an important research topic in data assimilation. In this paper, a novel parameter estimation method is introduced, where the parameter is considered as the state process in a nonlinear filtering problem and the state model that contains the parameter is used to construct a pseudo-observation. This approach is named the direct filter method since nonlinear filtering algorithms are used to estimate the parameter directly without estimating the state model as part of the solution in the nonlinear filtering problem. Numerical experiments are carried out to examine the effectiveness and accuracy of the direct filter method.",

keywords = "Bayesian inference, Data assimilation, Nonlinear filtering problem, Parameter estimation, State-space model",

author = "Richard Archibald and Feng Bao and Xuemin Tu",

note = "Publisher Copyright: {\textcopyright} 2019 Elsevier Inc.",

year = "2019",

month = dec,

day = "1",

doi = "10.1016/j.jcp.2019.108871",

language = "English",

volume = "398",

journal = "Journal of Computational Physics",

issn = "0021-9991",

publisher = "Elsevier",

}

TY - JOUR

T1 - A direct filter method for parameter estimation

AU - Archibald, Richard

AU - Bao, Feng

AU - Tu, Xuemin

PY - 2019/12/1

Y1 - 2019/12/1

N2 - Parameter estimation is an important research topic in data assimilation. In this paper, a novel parameter estimation method is introduced, where the parameter is considered as the state process in a nonlinear filtering problem and the state model that contains the parameter is used to construct a pseudo-observation. This approach is named the direct filter method since nonlinear filtering algorithms are used to estimate the parameter directly without estimating the state model as part of the solution in the nonlinear filtering problem. Numerical experiments are carried out to examine the effectiveness and accuracy of the direct filter method.

AB - Parameter estimation is an important research topic in data assimilation. In this paper, a novel parameter estimation method is introduced, where the parameter is considered as the state process in a nonlinear filtering problem and the state model that contains the parameter is used to construct a pseudo-observation. This approach is named the direct filter method since nonlinear filtering algorithms are used to estimate the parameter directly without estimating the state model as part of the solution in the nonlinear filtering problem. Numerical experiments are carried out to examine the effectiveness and accuracy of the direct filter method.

KW - Bayesian inference

KW - Data assimilation

KW - Nonlinear filtering problem

KW - Parameter estimation

KW - State-space model

UR - http://www.scopus.com/inward/record.url?scp=85070896545&partnerID=8YFLogxK

U2 - 10.1016/j.jcp.2019.108871

DO - 10.1016/j.jcp.2019.108871

M3 - Article

AN - SCOPUS:85070896545

SN - 0021-9991

VL - 398

JO - Journal of Computational Physics

JF - Journal of Computational Physics

M1 - 108871

ER -

A direct filter method for parameter estimation

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this