
This section lists optimization and linear algebra packages
interfaced with CUTEr and how they may be obtained. For more
detailed information on each package, please refer to the
Documentation section and in
particular the ManPages.
 ALGENCAN (part of the TANGO project)

a Fortran code for general nonlinear programming that does not use
matrix manipulations at all and, so, is able to solve extremely large
problems with moderate computer time. The general algorithm is of the Augmented
Lagrangian type and the subproblems are solved using GENCAN. GENCAN (included
in ALGENCAN) is a Fortran code for minimizing a smooth function with a
potentially large number of variables and boxconstraints.

 CG+ (Guanghui Liu, Jorge Nocedal and Richard Waltz)

a Conjugate Gradient code for solving largescale, unconstrained, nonlinear optimization problems. CG+ implements the FletcherReeves method, the PolakRibiere method, and the positive PolakRibiere method.

 CG Descent (William Hager and Hongchao Zhang)

A Conjugate Gradient method with guaranteed descent.

 Cobyla (Mike Powell)
 minimizes a nonlinear function subject to smooth nonlinear
constraints, using local linear approximations to the
constraints. There also exists a
Fortran 90/95 version
by Alan Miller.

 filterSQP (Roger Flecther,
Sven Leyffer)
 solves general nonlinear programs using SQP approximations at
each step and promotes convergence using a filter;

 HRB (Nick Gould)
 convert matrices derived from SIF problem data into HarwellBoeing
or RutherfordBoeing sparse formats. This packages comes standard with
the CUTEr distribution.

 Ipopt (Andreas Wächter)
 solves general nonlinear programs using a primaldual
interiorpoint approach using linesearches and filters to promote
global convergence.

 Knitro (Ziena Optimization Inc.)
 solves general nonlinear programs using a primaldual
interiorpoint approach globalized by trust
regions. Timelimited precompiled Knitro libraries may be
obtained from
Richard Waltz.

 LA04 (John Reid)
 solves linear programs using a steppestedge simplex method. This
package is part of the
HSL 2002
Fortran 90/95 subroutine library.

 LBFGS (Jorge Nocedal)

A limitedmemory quasiNewton code for largescale unconstrained
optimization.

 LBFGSB (Cyiou Zhou, Richard Byrd and Jorge Nocedal)
 a limitedmemory quasiNewton code for largescale
boundconstrained or unconstrained optimization.

 Loqo (Bob Vanderbei)
 solves general nonlinear programs using an infeasible
primaldual predictorcorrector approach. Timelimited
precompiled libraries for a variety of platforms are available
on the website.

 Minos (Bruce
Murtagh and Michael
Saunders)
 solves general nonlinear programs, preferably with sparse
constraints, by linearizing the nonlinear contributions so
efficient LP techniques may be applied;


 Nitsol (Michael Pernice and Homer Walker)
 solves systems of nonlinear equations using a globalized Newton iterative
method. Restarted GMRES is used to solve the Newton equations, is terminated as soon
as an inexact solution is found, and is safeguarded using a backtracking linesearch.

 Npsol (Philip Gill,
Walter Murray,
Michael Saunders and
Margaret
Wright)
 solves smooth constrained nonlinear programs using a dense SQP
algorithm and an augmentedLagrangian merit function.

 OSL (IBM )
 a library of standalong solvers for linear, quadratic and
mixedinteger programming. 60 days trial version available on the
website.

 PDS (Virginia Torczon)
 a collection of Fortran subroutines for solving unconstrained
nonlinear optimization problems using direct search methods.

 Praxis (John P. Chandler, Richard P. Brent)
 solves unconstrained problems without using derivatives. Praxis is
available in the opt section
of Netlib.

 Snopt (Philip Gill, Walter Murray and Michael Saunders)
 solves general nonlinear programs using a sparse SQP
algorithm with limitedmemory quasiNewton approximations to
the Hessian;

 Stenmin (Ali Bouaricha)
 a software package for large, sparse unconstrained optimization
using tensor methods. This package applies the methodology of TENMIN to
larger, sparse, problems.

 TAO (Steve Benson,
Lois Curfman McInnes,
Jorge Moré and
Jason Sarich)
 the Toolkit for Advanced Optimization focuses on the design and
implementation of componentbased optimization software for the
solution of largescale optimization applications.

 Tenmin (Robert B. Schnabel and T.T. Chow)
 solves small unconstrained minimization problems by either a tensor
method, using fourthorder models, or a standard method based on
quadratic models.

 TRON (ChihJen Lin and Jorge Moré)

A trust region Newton method for the solution of large boundconstrained optimization problems.

 Uncmin (Robert
B. Schnabel, John Koonatz and Barry E. Weiss)
 A modular system of algorithms for unconstrained minimization,
based on Newton or quasiNewton approaches using exact or inexact
second derivatives.

 VA15 (Jorge Nocedal)
 a subroutine for the solution of the largescale
unconstrained minimization problem. The routine does not require
(nor exploits) knowledge about the sparsity structure of the Hessian
matrix. It uses a limited memory BFGS quasiNewton method. This package
is available free of charge from the HSL archive.

 VE09 (Nick Gould)
 solves general, large, quadratic programs using an activeset
approach. It is available from the
HSL
subject to certain license agreements.

 VE12 (Nick Gould)
 solves general, large, quadratic programs using an interiorpoint
approach. It is available from the
HSL
subject to certain license agreements.

 VE14 (Nick Gould)
 solves general, large, quadratic programs using an interiorpoint
approach, using either a logarithmic barrier, the JittorntrumOsborne
barrier or the Lagrangian barrier. It is available from the
HSL
subject to certain license agreements.

 VF13 (Mike Powell)
 solves general nonlinear programs using an SQP approach, and
allows for nonmonotonic decrease of the merit function using
safeguards. This package is available free of charge from the
HSL archive.

