From: Vishy VisweswaranSubject: cGOP - A New Package for Global Optimization To: neum@cma.univie.ac.at Date: Wed, 20 Mar 1996 20:23:59 -0500 (EST) --------------------------------------------- cGOP -- A NEW PACKAGE FOR GLOBAL OPTIMIZATION --------------------------------------------- We are pleased to announce the availability of cGOP, a new software package for solving structured global optimization problems. A description of the package follows, along with instructions on how to retrieve, install and use the package. Dr. V. Visweswaran (Email: vishy@titan.princeton.edu) Professor C. A. Floudas (Email: floudas@titan.princeton.edu) Computer-Aided Systems Laboratory (WWW: http://titan.princeton.edu) Department of Chemical Engineering Princeton, NJ 08544 DESCRIPTION OF THE PACKAGE: -------------------------- cGOP is a package for rigorously solving nonconvex optimization problems to epsilon-global optimality. The package implements the GOP algorithm (Floudas and Visweswaran, 1993) which is a primal-dual decomposition algorithm applicable to general constrained biconvex problems, using a set of C subroutines to solve these problems using decomposition and branch-and-bound techniques. It also incorporates several improvements made to the original GOP algorithm to reduce the computational complexity (Visweswaran and Floudas, 1993) as well as new formulations that permit implicit solutions of some of the subproblems encountered during the algorithmic steps (Visweswaran and Floudas, 1996). The algorithms use local optimization solvers (currently MINOS 5.4 and CPLEX 3.0) to solve linear, mixed-integer linear and convex subproblems. This version of the cGOP package is a beta-test version, and can be used to solve problems with linear constraints. Problems can be specified using data files written in a high-level parsed language that permits their solution without re-compilation, and also using fixed-format data files. In addition, the package can be used via calls to C subroutines. cGOP has been in use at the Computer-Aided Systems Laboratory in Princeton University for the last two years, and has been used to solve problems involving several hundred variables and constraints. Currently, versions are available for Hewlett-Packard, Silicon Graphics and IBM RS6000 machine architectures. This is a beta-test version of the software, and is being released for use by academic researchers. A full version with the ability to solve nonlinearly constrained problems will be made available in the future. Commercial sites wishing to use this product should contact one of the authors directly. HOW TO OBTAIN THE SOFTWARE: -------------------------- The cGOP package is available through anonymous FTP from titan.princeton.edu. To retrieve the current version, perform the following steps: (1) Login in via anonymous ftp to titan.princeton.edu, using the user name "anonymous" and giving your email address as the password. The package is located in "pub/cGOP", and has three components - (a) cGOP.tar.Z, which contains the main libraries and examples, (b) cGOPdoc.ps.Z, which contains documentation for cGOP, and (c) cparsedoc.ps.Z, documentation for cparse, a high level parser that can be used to enter the problems for cGOP in an easy format. Make sure to set file type to binary before retrieving the files. Below, we show a sample session of using anonymous ftp: unix> ftp titan.princeton.edu Name: anonymous Password: yourname@yourmachine.yourdomain ftp> cd pub/cGOP ftp> type binary ftp> get cGOP.tar.Z ftp> get cGOPdoc.ps.Z ftp> get cparsedoc.ps.Z ftp> quit unix> (2) Select a directory where you want to store the distribution. Go to that directory, then: prompt> uncompress cGOP.tar.Z prompt> tar -xf cGOP.tar This will create a subdirectory called "Gop" under the current directory, and put the distribution there. (3) Uncompress and print out the manuals, cGOPdoc.ps and cparsedoc.ps. These contain full instructions for installing and using cGOP and the parser. (4) Look at the directories "cparse", "examples" and "callable". These contain example files that show how cGOP can be used through data files or using calls to C subroutines. (5) If you do retrieve the package, please send us an e-mail at cgop@titan.princeton.edu with a brief description of the machine you intend to run it on. This will help us to keep you informed of updates, bug fixes etc. If you have any difficulty getting the files, please send us an e-mail at cgop@titan.princeton.edu and we will be happy to assist you in obtaining the software.