Writing Global Optimization Software

Abstract

Global Optimization software packages for solving Mixed-Integer Non-linearOptimization Problems are usually complex pieces of codes. Some ofthe difficulties involved in coding a good GO software are: embeddingthird-party local optimization codes within the main global optimizationalgorithm; providing efficient memory representations of the optimizationproblem; making sure that every part of the code is fully reentrant.Finding good software engineering solutions for these difficultiesis not enough to make sure that the outcome will be a GO softwarethat works well. However, starting from a sound software design makesit easy to concentrate on improving the efficiency of the globaloptimization algorithm implementation. In this paper we discuss themain issues that arise when writing a global optimization softwarepackage, namely software architecture and design, symbolic manipulationof mathematical expressions, choice of local solvers and implementationof global solvers.