A weak method in numerical identification
The following problem is formulated in the paper: given the values of a function \(U\) in the linear envelope \(L\) of \(M\), find \(N\); \(a_1,a_2,\dots, a_N\in \mathbb{R}\); \(\lambda_1,\lambda_2,\dots, \lambda_N\in\Omega\) so that \(u(t)= \sum^N_{i=1} a_ie(\lambda_i,t)\), \(t\in[0,T]= I\), where \(\Omega\) is an open bounded set of \(\mathbb{R}^n\), \(e\in C(\overline\Omega\times I)\) such that the family \(M= \{e(\lambda,s),\lambda\in \Omega\}\) is linearly independent. To this problem, the author associates an equivalent weak formulation of the problem of the form: find \(\ell\in D\) such that \(A\ell= U\), where \(A:E'\to F\), \(E'\) the dual space of \(E= C(\overline\Omega)\), \(F= C(I)\), \(D= A^1(L)\subset E'\) is the linear envelope of the family \(\{\delta_\lambda, \lambda\in\Omega\}\), \(\delta_\lambda\) the Dirac measure at \(t\) in point \(\lambda\). The identification problem of \(N\), \(a_i\) and \(\lambda_i\) is equivalent with the determination of the distribution \(\ell= \sum^N_{i=1} a_i\delta_{\lambda_i}\). The method introduced by the author consists in the weak approximate determination of \(\ell\) in the weak topology metric sense in \(E'\). The proposed method is applied to the identification problem of a sum of exponentials and numerical results for \(N= 14,20,23\) and different test functions are given. Finally, the advantages of the proposed method in comparison with other methods are underlined.
- Development of integro-differential methods of parameter determination for a sum of exponential functions
- scientific article; zbMATH DE number 4147937
- Constructing accurate linear approximations to a class of exponential interpolation problems
- On some algorithms for non-parametric identification of linear systems
This page was built for publication: A weak method in numerical identification
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4269846)