Keynote Speakers


Michael Affenzeller, 

University of Applied Sciences of Upper Austria,
Hagenberg, Austria


"Real-World Systems Modeling and Optimization with Hybrid Evolutionary Algorithms"


Abstract:
The focus of the talk will be on systems modeling, optimization, algorithm design and analysis with respect to real world problems using different flavors of evolutionary algorithms. A special focus will be on the analysis of algorithm dynamics in order to analyze the reasons of premature convergence which provides deeper insights about the algorithms internal behavior and gives inspiration for the design of new hybrid algorithms. The talk will cover algorithm analysis, new variants of hybrid evolutionary algorithms like offspring selection or SASEGASA and its application in the field of combinatorial optimization, Genetic Programming-based symbolic regression and classification as well as simulation based optimization. Examples of real-world applications mainly from the field of industrial applications will exemplarily show the potential of hybrid evolutionary algorithms in different mainly industrial domains. As the open source optimization framework HeuristicLab http://dev.heuristiclab.com/ denotes the basis for all the theoretical as well as practical aspects covered in the talk, the potential of a generic framework for algorithm design, analysis and application will be one of the major inputs of the talk.
 






Mikko Kolehmainen,

University of Eastern Finland,
Kuopio, Finland


"Directions for systems modeling research"


Abstract:
The computational methods are in the forefront of the Big Data revolution concerning especially predictive modeling. Genetic Algorithms and Deep Neural Networks are valuable tools for these activities and their usage has provided new level of accuracy for regression and classification. However, these tools are inherently static and thus mostly suitable for interpolation within the data set used. Therefore, it is suggested that the focus of the research should be shifted to systems modeling which makes it possible to extrapolate i.e. answer what if questions not covered by the dataset used for training. Moreover, it is suggested that biological problems are of high importance as application areas for this kind of dynamic modeling. Two such areas namely health status prediction and aquatic modeling are presented as examples together with first results achieved. Additionally, the interconnections between systems modeling, static modeling and causality detection is discussed.