by Prof. Michael Affenzeller (Vice dean for R&D at University of Applied Sciences Upper Austria, Head of Studies for the Master Degree Program in Software Engineering, Head of the Heuristic and Evolutionary Algorithms Laboratory).

In modern computer science, there are numerous problems which are solved by a combination of modeling and simulation on the one hand and optimization by evolutionary algorithms on the other hand. For instance, in simulation based optimization, simulations can be used as evaluators for potential solutions, especially when these solution candidates are too complex to be formulated as compact formulations. Furthermore, there are often optimization tasks within a simulation that need to be optimized during the simulation. In this context, the computational power of infrastructures providing massively parallel high performance computing and / or cloud computing acts as an incubator for transforming ideas, that were initially designed for toy problems, to real world problem situations.

However, the obvious task of simulation based optimization is not the only point of interaction between these two fields of computer science. Data-based modeling techniques from the field of machine learning show the potential to establish surrogate models trained using data generated by simulators in the offline phase, which can for example act as ad-hoc estimators in simulation based optimization on a strategic level.

This presentation will cover theoretical aspects as well as real world examples demonstrating how the open source framework HeuristicLab can be used for modeling, optimization and machine learning tasks for concrete challenges in the domain of production, logistics and systems research.