From Free Software Directory
Jump to: navigation, search


A solver of large-scale unconstrained minimization problems.

The routine M1QN3 has been designed to minimize functions depending on a very large number of variables (several hundred million is sometimes possible) not subject to constraints. It implements a limited memory quasi-Newton technique (the L-BFGS method of J. Nocedal) with a dynamically updated scalar or diagonal preconditioner. It uses line-search to enforce global convergence; more precisely, the step-size is determined by the Fletcher-Lemaréchal algorithm and realizes the Wolfe conditions.




Download version 3.2 (mature)
released on 14 January 2009



LicenseVerified byVerified onNotes
License:GPLv3Kelly Hopkins14 January 2009

Leaders and contributors

Claude Lemarechal Maintainer
Jean-Charles Gilbert Maintainer

Resources and communication

AudienceResource typeURI

Software prerequisites

This entry (in part or in whole) was last reviewed on 14 February 2017.


Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.3 or any later version published by the Free Software Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license is included in the page “GNU Free Documentation License”.

The copyright and license notices on this page only apply to the text on this page. Any software or copyright-licenses or other similar notices described in this text has its own copyright notice and license, which can usually be found in the distribution or license text itself.