A new derivative free method for optimal experimental design utilizing a maximum a posteriori update

Author

Kawohl, M. ; Heine, T.

Author_Institution

Meas. & Control Group, Tech. Univ. Berlin, Berlin, Germany

fYear

2007

fDate

2-5 July 2007

Firstpage

5202

Lastpage

5207

Abstract

We present a new derivative free method for the calculation of the information content of a planned experiment in optimal experimental design (OED). It is shown that in case of a mathematical model that is linear in the model parameters the new approach yields the same result as a calculation with the Fisher Information Matrix (FIM). The new algorithm is derived from a maximum a posteriori (MAP) update. The needed statistical moments of the planned trajectory are calculated using a derivative free approach, the Unscented Transformation (UT). The algorithm is demonstrated using a simple, but very instructive biological growth model.

Keywords

covariance matrices; design of experiments; maximum likelihood estimation; optimisation; FIM; Fisher information matrix; MAP update; OED; UT; biological growth model; covariance matrix; derivative free method; information content; mathematical model; maximum a posteriori update; model parameters; optimal experimental design; planned experiment; planned trajectory; statistical moments; unscented transformation; Approximation methods; Biological system modeling; Covariance matrices; Mathematical model; Maximum likelihood estimation; Substrates; Trajectory; Maximum a posteriori update; Optimal experimental design; Unscented Transformation;

fLanguage

English

Publisher

ieee

Conference_Titel

Control Conference (ECC), 2007 European

Conference_Location

Kos

Print_ISBN

978-3-9524173-8-6

Type

conf

Filename

7068845