An Object Oriented Framework for Parallel Multiple Optimizations

Abstract

This report presents a general Objected-Oriented Framework for Parallel and Multiple Optimizations. We designed a general Objected-Oriented Class Library for Parallel Numerical Optimization. All optimization problems are considered: Unconstrained, Constrained, availability (or not) of Derivatives , availability (or not) of Second Derivatives. We also consider all type of methods for solving those problems: non-derivative methods, gradients methods, global optimisation methods. The report describes the C++ classes dealing with (almost) all possibilities of problems types and methods. This enables to have a very simple and efficient interface which will allow any user to easly define his problem, and to add any new optimization solver to his general framework. Examples using the framework in real-life for solving practical problems such as MBS (MultiBody Systems), are shown.