Sensitivity Analysis for Regularized Estimation in Some System Identification Problems

From MaRDI portal





Let \(A(u,\theta)=f\) be a functional equation where \(\theta\) has to be identified from measured information about \(u\) (\(f\) is taken to be known). The author considers a vector of measurements \(z\) such that \(z=Xu+\varepsilon\) where \(X\) is a linear mapping and \(\varepsilon\) a vector of measurement or possible modeling errors. The estimation problem for \(\theta\) can be put in the form of a nonlinear nonparametric regression. The situation is illustrated by some models described by partial differential equations (parabolic diffusion equation, hyperbolic wave equation). An approach to sensitivity calculation that may be used with regularization estimators is described. The Gauss-Newton algorithm is used for solving the identification problem and a concrete problem is numerically solved and the error is estimated by sensitivity analysis.











This page was built for publication: Sensitivity Analysis for Regularized Estimation in Some System Identification Problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3982104)