The Levenberg-Marquardt method: an overview of modern convergence theories and more

Research output: Contribution to journalResearch articleContributedpeer-review

Contributors

Abstract

The Levenberg–Marquardt method is a fundamental regularization technique for the Newton method applied to nonlinear equations, possibly constrained, and possibly with singular or even nonisolated solutions. We review the literature on the subject, in particular relating to each other various convergence frameworks and results. In this process, the analysis is performed from a unified perspective, and some new results are obtained as well. We discuss smooth and piecewise smooth equations, inexact solution of subproblems, and globalization techniques. Attention is also paid to the LP-Newton method, because of its relations to the Levenberg–Marquardt method.

Details

Original languageEnglish
Pages (from-to)33-67
Number of pages35
JournalComputational Optimization and Applications
Volume89
Issue number1
Publication statusPublished - Sept 2024
Peer-reviewedYes

External IDs

Scopus 85195660413

Keywords

Research priority areas of TU Dresden

DFG Classification of Subject Areas according to Review Boards

Subject groups, research areas, subject areas according to Destatis

Library keywords