PANEL SESSION 2 – Données et traitement automatisé – Data and Automated Treatment
(EN) Dr.-Ing. habil. Georg Krempl is assistant professor in the Algorithmic Data Analysis Group at Utrecht University in the Netherlands, and Privatdozent (entitled to supervise and promote PhD students in computer science) at Magdeburg University in Germany. His area of research is machine learning with focus on streaming and non-stationary, time-evolving data. For such data, he develops prediction methods that are capable to operate under limited supervision in dynamic environments. His research leads to a better understanding how environments change and which influence a machine learning algorithm has in this change, and helps to identify where such algorithms are uncertain or wrong in their predictions. In his habilitation thesis, defended 2016 at Magdeburg University, he introduced a family of new probabilistic active learning approaches that allow prediction systems to efficiently interact with human supervisors, by identifying samples with uncertain or difficult classification. During his Ph.D. research at Karl-Franzens-University Graz, he pioneered together with Vera Hofer research on predictive algorithms for applications with delayed supervision, i.e., where no immediate feedback on the predictions is available.
His research is applied in various areas, for example in credit scoring and default prediction in collaboration with tax and financial authorities; in medicine, for predicting the evolution in the recovery of brain trauma patients; and in Neurosciences, for actively tuning Brain-Computer Interfaces. He is regularly organising workshops, special sessions and tutorials at international machine learning and data science conferences.