Members
Jakub M. Tomczak
Associate professor(Group PI)
Babak Esmaeili
PostdocHaotian Chen
Ph.D. candidateMahdi Mehmanchi
Ph.D. candidateAnna Kuzina
Ph.D. candidate(External at Vrije Univ. Amsterdam)
Sharvaree Vadgama
Ph.D. candidate(External at the Univ. of Amsterdam)
Adam Izdebski
Ph.D. candidate(External at the Helmoltz Munich)
Jan Engelmann
Ph.D. candidate(External at the Helmoltz Munich)
In our research, we focus on developing new probabilistic models parameterized by deep neural networks.
We work on marginal generative models, conditional generative models and joint generative models, basing on the following frameworks:
- Variational Auto-Encoders
- Diffusion-based models
- Flow-based models
- Autoregressive models
- Energy-based models
- Score-based generative models
We apply our methods to various problems, e.g., Life Sciences, Molecular Sciences, signal processing, audio synthesis, image/video synthesis, text synthesis.
Our group is involved in teaching and supervision
To all prospective students: We are interested in the theoretical aspects of deep generative modeling, e.g., proposing new models and (preferably) theoretical analysis (e.g., formulating theorems, proving/showing properties). Applications of deep generative modeling are interesting as well, however, we must be aware of limited computational resources at the Univeristy. From our students we expect high independence (including proposing own ideas), good understanding of mathematics (algebra, calculus, statistics, probability theory) and good programming skills (Python + ML/DL libraries, preferably PyTorch). Please take a look at the template of a BSC/MSC thesis and get familiar with information therein.
Currently, we are involved in teaching the following course:
- Generative AI Models (TU/e)
The goal of our group is to share code together with our research
As a group, we truly believe in reproducible research. Therefore, we publish our code together with papers. You can check our repositories on our Github. If you cannot find a repository there, please visit github repositories of group members.
If any of our papers does not provide code, please contact Jakub.