Bayesian Inverse Problems

4

Parameter Identification Based on Conditional Expectation

 

Elmar Zander,^1,* Noémi Friedman² and Hermann G. Matthies¹

¹ Technische Universität Braunschweig, Germany.

² Institute for Computer Science and Control (SZTAKI), Budapest, Hungary.

^* Corresponding author: [e-mail]

Parameter identification is an important issue in many scientific and technical disciplines. We present here a technique that updates our knowledge of the system parameters based on the so-called conditional expectation. It will be shown that for linear systems with normally distributed uncertainties this reproduces exactly the Bayes’ posterior and is thus a non-linear extension of the Kalman filter.

 

4.1 Introduction

Estimation of model parameters is a very ...