scientific article; zbMATH DE number 4080732
[For the entire collection see Zbl 0643.00021.] Fitting with the \(\ell_ 1\)-norm is known to be more robust than the least-squares fit. Therefore, following \textit{P. J. Huber} [Ann. Stat. 1, 799-821 (1973; Zbl 0289.62033)], the authors consider the minimization of \(F(x)=\sum^{m}_{j1}\rho_ k(f_ j(x)/\sigma)\) where \(\rho_ k\) is the Huber function \(\rho_ k(t)=t^ 2/2\) if \(| t| \leq k\) and \(=k| t| -k^ 2/2\) otherwise. If the nonlinear functions \(f_ j\) are linearized, then the Gauss-Newton method is obtained. This is compared with other linearizations, e.g. the function \(\rho_ k\) may be also linearized and second order terms may be incorporated. Finally, the computation of a reasonable value of \(\sigma\) is discussed.
- Multiple neutral data fitting
- Fitting non-Gaussian persistent data
- scientific article; zbMATH DE number 3934293 (Why is no real title available?)
- scientific article; zbMATH DE number 4068146 (Why is no real title available?)
- scientific article; zbMATH DE number 1058093 (Why is no real title available?)
- scientific article; zbMATH DE number 3996989 (Why is no real title available?)
- scientific article; zbMATH DE number 795297 (Why is no real title available?)
- scientific article; zbMATH DE number 884961 (Why is no real title available?)
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3810784)