System Optimization

Due to many degrees of freedom, optimization tasks are often complex and build an enormously huge search space. It is mostly not possible to cover all major parts within all dimensions of the universe of theoretically possible solutions.

Artificial Intelligence (AI) and standard optimization techniques help to traverse such huge search space in a more efficient manner than the brute force search. However, these techniques also tend to increase the search space and the degrees of freedom, which need to be manually controlled. How powerful the AI techniques are, depends on how they are applied and integrated into the controlling expert systems.

Many competitors proudly focus on just a single AI technique and exploit it to the maximum possible curve fitting. On the contrary, we developed and extended our system optimization solutions to:

• Clearly model the optimization problem
• Target to flexibly combine the advantages of several AI techniques in an organized and extensible manner
• Select suitable hybrid AI modeling and optimization techniques
• Traverse through the search space, following sequences, paths and hierarchies
• Give the researcher the ability to model corresponding hierarchies and process flows and to optimize systems in several predefined steps towards same or different targets
• Use feedback and results from past optimization experiences in new optimization processes
• Dynamically filter unwanted parts of the search space
• Reduce the probability of curve fitting to increase the robustness

Our present implementations of Artificial Intelligence modules include standard and custom variations of:

• Fuzzy Logic
• Genetic Algorithms
• Artificial Neural Networks
• Neuro-Fuzzy Systems
• Support Vector Machines
• Differential Evolution
• Ant Colony Optimization
• Gene Expression Programming
• Singular Value Decomposition