Elsevier Science Home
Computer Physics Communications Program Library
Full text online from Science Direct
Programs in Physics & Physical Chemistry
CPC Home

[Licence| Download | New Version Template] aaxw_v3_0.tar.gz(517 Kbytes)
Manuscript Title: MERLIN-3.0, A multidimensional optimization environment.
Authors: D.G. Papageorgiou, I.N. Demetropoulos, I.E. Lagaris
Program title: MERLIN-3.0
Catalogue identifier: AAXW_v3_0
Distribution format: tar.gz
Journal reference: Comput. Phys. Commun. 109(1998)227
Programming language: Fortran.
Computer: Sun SPARCstation 5, Sun Classic, Sun Ultra-2, SGI Challenge-M, CD4680, Intel-based PCs, Macintosh.
Operating system: SunOS 4.1.3C/5.5.1, IRIX 6.2, EP/IX 1.4.3, Linux.
Word size: 32
Classification: 4.9.

Subprograms used:
Cat Id Title Reference
ABHA_v2_0 MCL-3.0 CPC 109(1998)250

Nature of problem:
Many problems in Physics, Chemistry, Engineering and in other disciplines are frequently reduced to minimizing a function of many variables. As examples we refer to systems of non-linear equations, to modeling, to variational methods, to curve fitting and to the training of neural networks.

Solution method:
MERLIN provides a programmable environment that makes the whole process of minimizing multidimensional functions with bound constraints, flexible and efficient. Ten algorithms and a strategy are implemented. Two of them use only function values, while the rest use gradient information as well. One algorithm is specific for functions that can be cast in a sum of squares form.

The only restriction is set by the available memory of the hardware configuration.

Unusual features:
The source code can be customized in regard to the required precision (single or double), to the maximum number of variables and to the maximum number of the squared terms, via a provided installer program. MERLIN can be easily extended by the user through a predesigned plug-in mechanism. Additional documentation is provided in the user manual (147 pages) that accompanies the distributed program.

Running time:
Depending on the objective function. The test run took 1.55 sec on a Sun SPARCstation 5.