Large Scale Parameter Meta-Optimization of Metaheuristic Optimization Algorithms with HeuristicLab Hive

In the recent decades many different metaheuristic algorithms have been developed and applied to various problems. According to the \textit{no free lunch} theorem no single algorithm exists that can solve all problems better than all other algorithms. This is one of the reasons why metaheuristic algorithms often have parameters which allow them to change their behavior in a certain range. However, finding good parameter values is not trivial and requires human expertise as well as time. The search for optimal parameter values can be seen as an optimization problem itself which can be solved by a metaheuristic optimization algorithm (\textit{meta-optimization}). In this paper the authors present the meta-optimization implementation for the heuristic optimization environment HeuristicLab. Because meta-optimization is extremely runtime intensive, a distributed computation infrastructure, HeuristicLab Hive, is used and will be described in this paper as well. To demonstrate the effectiveness of the implementation, a number of parameter optimization experiments are performed and analyzed.