Physics-Informed Machine Learning

"Physics-Informed" Machine Learning supplies common learning models more information than just data at their training time. Imagine a neural net, having access to differential equations and expert domain knowledge. This access can be e.g. done by including mathematical formulas in the learning architecture.

You will find more information, currently only in German language, on a dedicated page devoted to my book and my lectures in data mining and machine learning:

Multi-Agent Systems

Multi-Agent systems are an approach for solving distributed problems. They can be used to model social behaviour, game theoretical moves, markets and negotiations or machinery and process routes.

Evolutionary algorithms

Although sometimes not being the most optimal solution, evolutionary algorithms have some striking advantages that can be exploited for industrial processes. In this context, my research focuses on coordinated evolutionary strategies of agent coalitions or holons.