Dr. Oliver Grothe

Holder of Chair Analytics and Statistics at Kalrsruhe Institute of Technology

Conference theme:

Introduction to copulas with applications in machine learning and data preprocessing

Abstract:

Any multivariate distribution can be decomposed into its marginal distributions and its dependence structure, where the dependence structure is represented by a copula function. In this talk, we provide an introduction to copula modeling, its theoretical foundations and links to current challenges in machine learning. By doing so, we want to emphasize the importance of this statistical tool to researchers in machine learning and information science. We point out the close relationship to concepts familiar to these researchers, such as mutual information or entropy. Taking the perspective of copula modeling means identifying patterns, structures, or associations in data without being distracted by the scale of single variables. Throughout the talk, we give examples of promising applications of copulas in areas such as generative models, representation learning, data augmentation, Bayesian network modeling, and anomaly detection.

Speaker biography:

Oliver Grothe is full professor for analytics and statistics and dean of studies for mathematical economics at the Karlsruhe Institute of Technology (KIT). He received his doctorate in 2008 from the University of Cologne as part of the Research Training Group for Risk Management and was awarded the venia legendi in statistics and econometrics in 2013. He gained insights into issues relevant to practice through his consulting work for energy suppliers. Oliver Grothe has numerous, interdisciplinary research interests between analytics, statistics, and machine learning. His particular areas of focus include dependence structures in high-dimensional data, networks and processes, risk measurement and financial market modeling, prediction and optimal design in energy markets, and problems in medical statistics.