Michael Lohaus

In October 2017 I started my PhD supervised by Ulrike von Luxburg in the Theory of Machine Learning group (TML) at Tuebingen university, computer science department. I am a scholar in the International Max Planck Research School for Intelligent Systems (IMPRS-IS), a graduate school for PhD students from both university and Max-Planck-Institute in Tuebingen and Stuttgart.

In 2011 I started studying Mathematics and Economics at Cologne University. During my Bachelor and Master programs I visited Lund University in Sweden and Eötvös Loránd University in Budapest for one semester. In 2017 I finished my Master thesis on a Primal-Dual Algorithm for Semidefinite Programs applied on the MaxCut SDP relaxation.

At the beginning of my PhD, I worked on ordinal data which can be used for classic machine learning problems. In ordinal data we are not given actual distances between data points, but we know if a distance from A to B is smaller than the distance from A to C. I started working on ordinal embeddings, triplet prediction problems or learning the underlying metric between those data points.

Secondly, I am interested in the real-world impact machine learning has on society. In particular, I study classification problems under fairness considerations. In my recent work, I consider a setting where each point belongs to one of two sensitive groups, and we want the result of a decision to be fair in a certain sense with respect to the two groups.

• Michael Lohaus, Michaël Perrot, Ulrike von Luxburg, Too Relaxed to Be Fair. In The 37th International Conference on Machine Learning (ICML 2020) pdf

• Michael Lohaus, Philipp Hennig, Ulrike von Luxburg, Uncertainty Estimates for Ordinal Embeddings. arXiv