Screenshots

Optimization Tool Box

An example of the different perfomances using Genetic Algorithms (black) and Evoution Strategies (red) on the F1 benchmark function using different mutation operators and selection methods.

Application Viewers

This is an example for multi-objective optimization in financial engineering using Multi-Objecitve Evolutionary Algorihtms. Using the Markowitz mean-variance model the Portfolio Selection Problem is given as to maximize the expected return (y axis) and to minimize the expected risk (x axis) by distributing your investment over multiple alternative assets. In this problem instance we applied a cardinality constraint of k = 2, thus limiting the number of assets in the portfolio to two. The graphic gives the optained pareto-front (red) and the current solutions in the population (blue crosses) compared to the pareto-front of the unconstrainted Portfolio Selection Problem (red) and the available assets (black crosses).

The Artificial Ant problem is given as to find a search strategy for an artificial ant to find as many food particles as possible with a limited amount of steps (=400). The ant is living in a torodial world where the food particles are distributed along the so called Sant-Fee trail as black dots. The ant starts in the upper left corner of the 2D projection of the torodial world and is controlled by a program that is to be evolved. The resulting path of the ant is colour coded. At the beginning of the simulation the path is light green and turns into deep purple toward the end of the simulation. Every food particle that the ant collected is coloured red uncollected food particles remain black. In the lower part of the problem frame you can the evolved program code that generated the displayed ant path.

This is an example of a Traveling Salesman Problem optimzed using a Genetic Algorithm. Here the task is given by identifing the shortes possible path visting all cities.