The research activities of the Intelligent Systems group are focused on machine learning, a scientific discipline in the intersection of computer science, statistics, and applied mathematics. The importance of this discipline has continuously increased in the recent past, and meanwhile, machine learning has developed into one of the core methodologies of modern artificial intelligence as well as the emerging research field of data science.
Beyond its use as a key component of intelligent systems design, machine learning and related fields such as data mining have a more far-reaching influence on science and society as a whole. In fact, statistical methods for learning from data and efficient algorithms for analyzing massive amounts of data provide the scientific and technological foundation of our "information society" in the era of "big data".
In recent years, we have been working on various topics within machine learning, including preference learning, multi-target prediction, online learning and data streams, reinforcement learning, and reliable prediction. More detailed information is provided in the description of our research projects and the list of publications.
Another branch of research in our group is centered around the notion of uncertainty in knowledge-based systems and related concepts such as imprecision and vagueness. In practice, knowledge is always afflicted with uncertainty and various sorts of imperfection, regardless of whether it has been expressed explicitly by a human expert or derived through machine learning or data mining techniques. Therefore, the question of how to represent and process uncertain information in the most appropriate way is of utmost importance for the design of intelligent and knowledge-based systems.
Without questioning the central role of probability theory as a mathematical tool for modeling and reasoning under uncertainty, we consider formalisms and uncertainty calculi such as evidence theory, possibility theory, and fuzzy set theory as viable alternatives. In other words, we doubt that all types of uncertainty can be captured appropriately in terms of a single formalism; instead, the best approach will depend on the concrete problem at hand.
Although the focus of our research is on theoretical questions and methodological problems, we are also very interested in practical applications of machine learning and data mining. In recent years, we have been working on such applications in various domains, notably in biology and the life sciences. Yet, we are also looking at practical problems in other fields, including engineering, economics, and the humanities.