From 2013 to 2018, I was pleased to work in the Empirical Inference department of the MPI for Intelligent Systems as a PhD student and as a postdoc. My research focused on causal inference and multi-agent systems as well as, more broadly, machine learning and time series analysis. While I mainly worked on mathematical theory and algorithmic implementations, my work also included large-scale experiments and stakeholder management, in particular w.r.t. social data privacy norms.
Here are some more specifics on my work at the MPI:
During my postdoc, I was leading the Cafeteria Coordination project (see the project page for details). There I was using machine learning, time series analysis and game theory to support coordination between users of congested facilities and achieve higher efficiency, together with colleagues and external collaborators. The project involves both, theory (check out our preprint), application/experiments (in the form of our cafeteria congestion forecast app which is unfortunately only accessible within the campus) and stakeholder management, in particular to account for social data privacy norms.
During my PhD studies, I worked on causal inference. Specifically, I worked on inference of causal models from time series data and quasi-experimental designs. Information theory often turned out to be a useful tool there. Furthermore, I applied causal models and machine learning methods like Gaussian processes to data-driven decision making problems in cloud computing and automatic intelligent agents (see my PhD thesis and the publications page for details). I was supervised by Bernhard Schölkopf, Dominik Janzing and Marc Toussaint.
In Proceedings of the 32nd International Conference on Machine Learning, 37, pages: 1898–1906, JMLR Workshop and Conference Proceedings, (Editors: F. Bach and D. Blei), JMLR, ICML, 2015 (inproceedings)
In Proceedings of the 32nd International Conference on Machine Learning, 37, pages: 1917–1925, JMLR Workshop and Conference Proceedings, (Editors: F. Bach and D. Blei), JMLR, ICML, 2015 (inproceedings)
Unser Ziel ist es, die Prinzipien von Wahrnehmen, Lernen und Handeln in autonomen Systemen zu verstehen, die mit komplexen Umgebungen interagieren. Das Verständnis wollen wir nutzen, um künstliche intelligente Systeme zu entwickeln.