I am a Ph.D. candidate at Uppsala University, Sweden advised by Thomas Schön (main), Niklas Wahlström and Antônio H. Ribeiro. I am fully supported by WASP.
In my research I try to understand deep learning through theory and empirical observations. I am interested in
- the double-descent phenomenon and the role of overparameterization
- training of neural networks and its implicit bias
- transfer learning and unsupervised pre-training
On the applied side, my research uses deep models for the evaluation of electrocardiogram (ECG) recordings to (1) help physicians with diagnosis and (2) explore the information within those recordings.
Background
- Summer 2024 will finish my Ph.D. at Uppsala University, Sweden
- July 2019 received my M.Sc. in systems and control from TU Delft, the Netherlands
- August 2015 received my B.Sc. in aerospace engineering from DHBW, Germany
- 1994 born
Latest research results and news
May 11, 2023, Accepted Paper: Our paper on invertible kernel PCA got accepted at IEEE Signal Processing Letters. We present a simple but effective method to invert the kernel PCA operations in an invertible subdomain which is sufficient for reconstruction in many scenarios.
doi arXiv code
January 03, 2023, Research Visit: I will visit Misha Belkin’s group at UC San Diego from 1st of March to 27th of May. During that time we will work on generalization of overparameterized models.
Misha Belkin
December 05, 2022, Poster Presentation: I presented a poster on our ongoing work in ECG-regression for electrolyte prediction at the annual AI4Reserach workshop, Uppsala.
poster workshop
November 22, 2022, Invited Talk: The bi-annual joint meeting from the Swedish and Danish societies for biopharmaceutical/medical statistics (DSBS, FMS), invited me to talk about our latest work in prediction models for myocardial infarctions from ECGs.
program slides
All news can be found here