A numerically efficient implementation of the expectation maximization algorithm for state space models

Wolfgang Mader*, Yannick Linke, Malenka Mader, Linda Sommerlade, Jens Timmer, Bjoern Schelter

*Corresponding author for this work

Research output: Contribution to journalArticle

9 Citations (Scopus)
3 Downloads (Pure)

Abstract

Empirical time series are subject to observational noise. Naive approaches that estimate parameters in stochastic models for such time series are likely to fail due to the error-in-variables challenge. State space models (SSM) explicitly include observational noise. Applying the expectation maximization (EM) algorithm together with the Kalman filter constitute a robust iterative procedure to estimate model parameters in the SSM as well as an approach to denoise the signal. The EM algorithm provides maximum likelihood parameter estimates at convergence. The drawback of this approach is its high computational demand. Here, we present an optimized implementation and demonstrate its superior performance to naive algorithms or implementations. (C) 2014 Elsevier Inc. All rights reserved.

Original languageEnglish
Pages (from-to)222-232
Number of pages11
JournalApplied Mathematics and Computation
Volume241
Early online date3 Jun 2014
DOIs
Publication statusPublished - 15 Aug 2014

Keywords

  • Kalman filter
  • expectation-maximization algorithm
  • parameter estimation
  • state-space model
  • maximum-likelihood
  • systems

Cite this

A numerically efficient implementation of the expectation maximization algorithm for state space models. / Mader, Wolfgang; Linke, Yannick; Mader, Malenka; Sommerlade, Linda; Timmer, Jens; Schelter, Bjoern.

In: Applied Mathematics and Computation, Vol. 241, 15.08.2014, p. 222-232.

Research output: Contribution to journalArticle

Mader, Wolfgang ; Linke, Yannick ; Mader, Malenka ; Sommerlade, Linda ; Timmer, Jens ; Schelter, Bjoern. / A numerically efficient implementation of the expectation maximization algorithm for state space models. In: Applied Mathematics and Computation. 2014 ; Vol. 241. pp. 222-232.
@article{adf6c229633643c6a65fa4c8f84602ef,
title = "A numerically efficient implementation of the expectation maximization algorithm for state space models",
abstract = "Empirical time series are subject to observational noise. Naive approaches that estimate parameters in stochastic models for such time series are likely to fail due to the error-in-variables challenge. State space models (SSM) explicitly include observational noise. Applying the expectation maximization (EM) algorithm together with the Kalman filter constitute a robust iterative procedure to estimate model parameters in the SSM as well as an approach to denoise the signal. The EM algorithm provides maximum likelihood parameter estimates at convergence. The drawback of this approach is its high computational demand. Here, we present an optimized implementation and demonstrate its superior performance to naive algorithms or implementations. (C) 2014 Elsevier Inc. All rights reserved.",
keywords = "Kalman filter, expectation-maximization algorithm, parameter estimation, state-space model, maximum-likelihood, systems",
author = "Wolfgang Mader and Yannick Linke and Malenka Mader and Linda Sommerlade and Jens Timmer and Bjoern Schelter",
year = "2014",
month = "8",
day = "15",
doi = "10.1016/j.amc.2014.05.021",
language = "English",
volume = "241",
pages = "222--232",
journal = "Applied Mathematics and Computation",
issn = "0096-3003",
publisher = "Elsevier",

}

TY - JOUR

T1 - A numerically efficient implementation of the expectation maximization algorithm for state space models

AU - Mader, Wolfgang

AU - Linke, Yannick

AU - Mader, Malenka

AU - Sommerlade, Linda

AU - Timmer, Jens

AU - Schelter, Bjoern

PY - 2014/8/15

Y1 - 2014/8/15

N2 - Empirical time series are subject to observational noise. Naive approaches that estimate parameters in stochastic models for such time series are likely to fail due to the error-in-variables challenge. State space models (SSM) explicitly include observational noise. Applying the expectation maximization (EM) algorithm together with the Kalman filter constitute a robust iterative procedure to estimate model parameters in the SSM as well as an approach to denoise the signal. The EM algorithm provides maximum likelihood parameter estimates at convergence. The drawback of this approach is its high computational demand. Here, we present an optimized implementation and demonstrate its superior performance to naive algorithms or implementations. (C) 2014 Elsevier Inc. All rights reserved.

AB - Empirical time series are subject to observational noise. Naive approaches that estimate parameters in stochastic models for such time series are likely to fail due to the error-in-variables challenge. State space models (SSM) explicitly include observational noise. Applying the expectation maximization (EM) algorithm together with the Kalman filter constitute a robust iterative procedure to estimate model parameters in the SSM as well as an approach to denoise the signal. The EM algorithm provides maximum likelihood parameter estimates at convergence. The drawback of this approach is its high computational demand. Here, we present an optimized implementation and demonstrate its superior performance to naive algorithms or implementations. (C) 2014 Elsevier Inc. All rights reserved.

KW - Kalman filter

KW - expectation-maximization algorithm

KW - parameter estimation

KW - state-space model

KW - maximum-likelihood

KW - systems

U2 - 10.1016/j.amc.2014.05.021

DO - 10.1016/j.amc.2014.05.021

M3 - Article

VL - 241

SP - 222

EP - 232

JO - Applied Mathematics and Computation

JF - Applied Mathematics and Computation

SN - 0096-3003

ER -