Since Sep. 2015, I am a PhD student in the Brain-State Decoding Lab at the University of Freiburg which is headed by Dr. Michael Tangermann. My primary research focus is on Brain-Computer Interfaces (BCis). These are systems which can translate neuronal signals (e.g. brain signals measured from the EEG) into control commands (e.g. to allow communication or enable wheelchair control for patients, or to enrich games) or for rehabilitation after stroke.
In BCI, my research is on two aspects. On the theoretical side, I work on unsupervised machine learning methods that can calibrate itself only by using only unlabeled data (data where the user's intention are unknown), hence, making a calibration session superfluous. On the practical side, I am involved in a project exploring the feasibility and effectiveness of a training based on Brain-Computer Interfaces for patients with language deficits (aphasia) after a stroke.
Before being a PhD student, I did a BSc. of Mathematics in Potsdam (Germany) from 2009-2012. After studying abroad for one semester in Perth (Australia) I did a MSc. of Applied Mathematics / Scientific Programming as part of a European program called COSSE (Computer Simulation for Science and Engineering) from 2013-2015. In this program, I spent the first year at TU Delft (Netherlands) and my second year at KTH Stockholm (Sweden) where I wrote my Master Thesis in computational neuroscience.
I used the gap year between my Bachelors and Masters to travel around Australia, New Zealand and South-East Asia. During that period, I first really got in contact with photography. Some of my pictures are shown on this homepage.
(Jan 2019): We gave a talk about our aphasia project at the applied machine learning days (AMLD 2019) - a highly inspiring event.
(Nov 2018): Here is a video how unsupervised machine learning can be used to rapidly allow a user to spell some letters only by using his brain signals and without prior calibration. Thanks to my cousin Daniel Knobloch for participating in this video.
(Sep 2018): A new article appeared in Frontiers in Human Neuroscience where we have compared the usage of auditory BCIs with open- and with closed-eyes.
Interestingly, the majority of subjects preferred to close their eyes while controlling the BCI – something which is not yet being used in practical applications.
This work is the result of our annual practical course where the students Albrecht Schall and Natalie Prange showed great effort. Link (open access)
(Apr 2018): We have been nominated for the BCI Award again: This time with our work on BCI-supported language training. Check out our video below. The winner will be crowned at the 7th International BCI Meeting May 21 – 25, 2018 at the Asilomar Conference Center in Pacific Grove, California, USA. There, we will also have a workshop on unsupervised learning.
Update: We won the 2nd price in the BCI award! Yay! News
(Apr 2018): Our new paper on unsupervised learning appeared in the IEEE computational intelligence magazine.
In this paper, we review different unsupervised learning approaches for ERP-based BCIs and compare three of them in an online study.
(Jan 2018): We presented a workshop about unsupervised learning for BCIs at the applied machine learning days in Lausanne – a great event where I would love to return next year Link
(Sep 2017): I won the best talk award for an oral presentation at the
7th Graz Brain-Computer Interface Conference 2017. I presented our joint work on unsupervised learning by mixing model estimators. Thanks again to the great collaborators.
(Aug 2017): Our work on unsupervised learning has been nominated for the BCI Award 2017,
an international award with more than 50 competing groups. A great collaboration with Pieter-Jan Kindermans, Thibault Verhoeven,
Klaus-Robert Müller and Michael Tangermann. Poster
(Jul 2017): I won the best talk award at the first Neuroadaptive Technology (NAT 17) conference in Berlin.
I presented work about unsupervised learning with learning from label proportions (LLP).
(Apr 2017): A new idea for BCIs: Learning from label proportions has been used in other areas as a simple and powerful concept to allow unsupervised learning. Together with Pieter-Jan Kindermans,
Thibault Verhoeven, Konstantin Schmid, Klaus-Robert Müller and Michael Tangermann, we demonstrate how this can be applied in ERP-based BCIs to obtain the first unsupervised classifier with guaranteed convergence.
If you are interested, then check out the project page or have a look at the article in PLOS ONE. Link (open access)