Automatic Column Scaling Strategies for Quasi-Newton Methods

Marucha Lalee, Jorge Nocedal

Research output: Contribution to journalArticlepeer-review

Abstract

A class of algorithms is described for unconstrained optimization, based on the BFGS update formula, which includes an automatic column scaling strategy. The new algorithms generalize the method of Powell [Math. Programming, 38 (1987), pp. 29–46]. Conditions are given on the scaling strategies that guarantee global and superlinear convergence on convex problems.
Original languageEnglish
Pages (from-to)637-653
JournalSIAM Journal on Optimization
Volume3
DOIs
StatePublished - 1994

Fingerprint

Dive into the research topics of 'Automatic Column Scaling Strategies for Quasi-Newton Methods'. Together they form a unique fingerprint.

Cite this