Modular Optimizer for Mixed Integer Programming MOMIP Version 1.1

Ogryczak, W. & Zorychta, K. (1993). Modular Optimizer for Mixed Integer Programming MOMIP Version 1.1. IIASA Working Paper. IIASA, Laxenburg, Austria: WP-93-055

Preview

Text WP-93-055.pdf Download (1MB) | Preview

Abstract

This Working Paper documents the Modular Optimizer for Mixed Integer Programming (MOMIP). MOMIP is an optimization solver for middle-size mixed integer programming problems, based on a modified branch-and-bound algorithm. It is designed as part of a wider linear programming modular library being developed within the IIASA CSA project on "Methodology and Techniques of Decision Analysis". The library is a collection of independent modules, implemented as C++ classes, providing all the necessary functions of data input, data transfer, problem solution, and results output. The Input/Output module provides data structure to store a problem and its solution in a standardized form as well as standard input and output functions. All the solver modules take the problem data from the Input/Output module and return the solutions to this module. Thus, for straightforward use, one can configure a simple optimization system using only the Input/Output module and an appropriate solver module. More complex analysis may require use of more than one solver module. Moreover, for complex analysis of real-life problems, it may be more convenient to incorporate the library modules into an application program. This will allow the user to proceed with direct feeding of the problem data generated in the program and direct withdrawal results for further analysis.

The paper provides the complete description of the MOMIP module. Methodological background allows the user to understand the implemented algorithm and efficient use of its control parameters for various analyses. The module description provides all the information necessary to make MOMIP operational. It is additionally illustrated with a tutorial example and a sample program. Modeling recommendations are also provided, explaining how to built mixed integer models in order to speedup the solution process. These may be interesting, not only for the MOMIP users, but also for users of any mixed integer programming software.

Actions (login required)

View Item