scientific article; zbMATH DE number 4060674

From MaRDI portal
Publication:3795178





In the first part of this paper the author reviews briefly some of the methods for the numerical evaluation of the Laplace inversion integral and cites related literature. The Laplace transform of f is denoted by g so \((*)\quad \int^{\infty}_{0}e^{-st}f(t)dt=g(s).\) Given g(s), \(s\geq 0\), the problem is to find f(t) numerically. By making the substitutions \(s=\alpha\) x and \(t=\alpha^{-y}\), \(\alpha >1\), the equation (*) is converted into the convolution equation \(\int^{\infty}_{-\infty}K(x-y)F(y)dy=G(x),\) where \(G(x)=sg(s),\) \(K(x)=(\log \alpha)se^{-s}\) and \(F(y)=f(t).\) It is assumed that G(x)\(\to 0\) as \(| x| \to \infty\). The author applies Tikhonov regularization to the numerical deconvolution problem using the operator \(Lf:=f\) for the smoothing so that the penalty term in the regularization functional is \(\lambda \| f\|\) 2. The maximum likelihood method is used for the determination of an optimal value for the regularization parameter. The given data consists of the values of \(g_ n\) at a fixed number N of equidistant points \(\{x_ n\}\), where it is assunmed that data are noisy: \(g_ n=u_ n+\epsilon_ n\), and both \(\{u_ n\}\) and \(\{\epsilon_ n\}\) are identified with independent stationary stochastic processes. Numerical results involving four test examples (from the iterature) are given and the numerical results are shown to be better than some of the methods that tested the same examples. Unfortunately this paper has a large number of typographical errors.











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 Q3795178)