Forschungsthemen
[MA] A Comparative Study of Genetic Optimization Approaches in Multi-Quality Auto-Tuning
The Software Technology Group has developed an approach that enables the model-driven development of self-optimising systems. This approach, Multi-Quality Auto-Tuning (MQuAT), is already implemented and uses a set of solvers to find optimal solutions. A common model of the abstract problem addressed by MQuAT is transformed into various alternative representations, which are understood by the solvers supported in the system; thus, model transformations are an essential part of the approach. So far, Integer Linear Programming, Pseudo-Boolean Optimisation, Ant Colony Optimisation, and a naïve, yet incrementally working local search algorithm have been used in MQuAT. The multitude of different solvers in combination with the huge amount of their respective configurations encourages the comparison of the solvers for different problem classes. Additionally, a huge and promising class of solvers has not been investigated so far: genetic algorithms. The goal of this thesis is to investigate the use of genetic algorithm solvers within MQuAT and to compare these to the solvers supported already.
Betreuer: Johannes Mey