and make the following selection:Multiobjective optimization
Let's find the optimal set of solutions for two functions whose minimum occur
for different design points. The functions we would be looking at are:
rotated ellipsoid and
moved ellipsoid. They are stored in
OptimizationModels.qsl. All participating function should have the same number
of variables. The boundaries for the first function will be initially applied.
Choose the Optimization icon from the toolbar
and make the following selection:
Genetic algorithm is currently the only available Multiobjective optimization method. It is based on the NSGA2 algorithm, developed by Prof. Deb Kalmonoy.
Click Optimize and wait until the optimization completes. When it does, the result is the so called 'Pareto front', which contains the non-dominated solutions - these are solutions that at least in one objective are better than the rest.
The number of points on the Pareto front, can be between 1 and the size of the GA population. The Pareto front is visualized with the brown square symbols:
Pareto front does not give a single solution to the optimization problem, but it provides the designer with a set of potentially optimal solutions. By studying the Pareto front, the designer can choose designs that suits his needs. By holding the CTRL key and clicking on a point, you can set a marker, and see where this points appears in the 'variable' or 'objective' space.
Detailed information about the values of each marker appears in the log screen:
This technique is applicable to any number of objectives.
See also