Identification de systèmes dynamiques linéaires à effets mixtes : applications aux dynamiques de populations cellulaires

Abstract : System identification is a data-driven input-output modeling approach more and more used in biology and biomedicine. In this application context, methods of experimental design are often used to test effects of qualitative factors on the response and each assay is always replicated to estimate the reproducibility of outcomes. The inference of the modeling conclusions to the whole population requires to account within the modeling procedure for the explained variability (fixed effects) and the unexplained variabilities (random effects) between the individual responses. One solution consists in using mixed effects models but up to now no similar approach exists in the system identification literature. The objective of this thesis is to fill this gap by using hierarchical model structures introducing mixed effects within polynomial black-box representations of linear dynamical systems. A new method is developed to estimate parameters of model structures such as ARX or Box-Jenkins. A solution is also proposed to compute the Fisher’s matrix. Finally, three application studies are carried out and emphasize the practical relevance of the proposed approach to identify populations of dynamical systems
Complete list of metadatas

https://hal.univ-lorraine.fr/tel-01755061
Contributor : Memoires Ul <>
Submitted on : Friday, March 30, 2018 - 12:08:22 PM
Last modification on : Friday, May 17, 2019 - 11:39:42 AM

Identifiers

  • HAL Id : tel-01755061, version 1

Citation

Levy Batista. Identification de systèmes dynamiques linéaires à effets mixtes : applications aux dynamiques de populations cellulaires. Automatique / Robotique. Université de Lorraine, 2017. Français. ⟨NNT : 2017LORR0224⟩. ⟨tel-01755061⟩

Share

Metrics

Record views

76