Skip to Main content Skip to Navigation
Conference papers

Bayesian noise model selection and system identification using Chib's approximation based on the Metropolis-Hastings sampler

Abstract : In this paper the problem of model selection has been applied to the identification of the model for the noise that affects the measured data. Performing the identification of a second order system has been considered as a side problem, in the view that the output of such a system is what we are measuring. A joint solution for the two distinct problems has been proposed in the context of the Bayesian statistical modelling. The main problem was with the approximation of the model evidence the solving of which required the use of numerical methods. However, our approach is not based on the well-known estimator of the harmonic mean which can exhibit bad convergence properties. As an alternative, the proposed estimation method takes advantage of the reversibility property of the Metropolis sub-kernel. The performances of the proposed solution have been assessed with encouraging results.
Document type :
Conference papers
Complete list of metadatas

Cited literature [10 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01729077
Contributor : Andrei-Cristian Barbos <>
Submitted on : Monday, March 12, 2018 - 12:31:43 PM
Last modification on : Tuesday, March 13, 2018 - 1:06:57 AM
Document(s) archivé(s) le : Wednesday, June 13, 2018 - 1:41:35 PM

File

cr_barbos336.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01729077, version 1

Citation

Andrei-Cristian Bărbos, Audrey Giremus, Jean-François Giovannelli. Bayesian noise model selection and system identification using Chib's approximation based on the Metropolis-Hastings sampler. XXVème Colloque GRETSI, Sep 2015, Lyon, France. ⟨hal-01729077⟩

Share

Metrics

Record views

71

Files downloads

39