Unknown
In:
Protokoll über das 24. Schmucker-Weidelt-Kolloquium für Elektromagnetische Tiefenforschung
Publication Date:
2020-02-12
Description:
Krylov subspace methods are well-known as iterative solvers for linear systems of equations. Using the same basic projection methodology these can also be used to approximate the solution of least squares problems, while avoiding the explicit construction of the sensitivity matrix. We point out how Krylov subspace methods such as CGNR – as well as its more robust variant LSQR – take into account the least squares origin of the normal equations, thereby ameliorating the ill-conditioning inherent in the inverse problem. These observations are illustrated with examples from the inversion of 3-D helicopter electromagnetic data. The research is carried out in an interdisciplinary project called ‘AIDA – From Airborne Data Inversion to In-Depth Analysis’.
Language:
English
Type:
info:eu-repo/semantics/conferenceObject
Format:
application/pdf
Permalink
|
Location |
Call Number |
Expected |
Availability |