GloptLab

Global Optimization Laboratory is an easy-to-use testing and development platform for solving quadratic constraint satisfaction problems, written in Matlab. All implemented methods are rigorous, hence it is guaranteed that no feasible point is lost. As the name suggests GloptLab will solve otimization problems but in the current release this feature is not yet available.

The project was funded by the Research Grant FSP 506/003 of the University of Vienna. The project is supervised by Prof. Arnold Neumaier, and the project assistants are Dr. Martin Fuchs and me.

DOWNLOAD GloptLab

The inputs are converted to a quadratic format, which forms the basis of the algorithms implemented in GloptLab, which are used to reduce the search space: scaling, constraint propagation, linear relaxations, strictly convex enclosures, conic methods, and branch and bound. From the method repertoire custom made strategies can be built, and the solution process can be viewed in a user friendly graphical interface:

GloptLab Screenshot

Building a solution strategy in the graphical user interface is done by inserting, editing or removing tasks (1). There is also a syntax check included, which prevents the creation of incorrect strategies. In the graphical user interface not only the strategies can be edited but a single problem or a problem list (2) can be solved by executing a strategy using the execute button. The text output of the solution procedure can be found in the text output window (3), while the graphical output for the problems in the graphical output window (4).

Important information of the currently selected problem (name, number of variables and constraints etc.) can be viewed in the right lower part (5). Creating new problems or converting existing ones in the GloptLab format is also possible with the conversion tools and the internal GloptLab editor. They can be accessed from the panel marked with (6). In the central long panel (7) the parameters for the graphical output, the statistic database, the automatically generated proofs, the profiler and the general configuration can be accessed and modified. The GloptLab configuration consists of several global parameters like the path of the external solvers or the width of a box which is assumed as tiny, and all the default values of the parameters used in the different task. There can be different configurations files, and the parameters contained can be edited by the user.


Related papers


Slides