Tabular data
Representations and predictive models for tabular data, including tabular foundation models and in-context learning, with missing values treated as part of the problem rather than a preprocessing step.
Machine learning
for real-world data
and explainable AI
We develop machine learning for data that comes from real scientific and industrial problems, and explanations that make a model's decisions legible to the people who act on them.
Part of GMUM, Jagiellonian University, Kraków
Research
Scientific data
Chemoinformatics and bioinformatics, where a prediction that is wrong or unintelligible carries a real cost, and where the data rarely looks like a benchmark.
Learning with few labels
Self-supervised and unsupervised representations, few-shot learning, density estimation and anomaly detection, for settings where annotation is expensive or simply unavailable.
Explainable AI
Counterfactual explanations, that is the smallest change to an input that flips a decision, and explanations expressed in text and concepts rather than in pixels or parameters.
People
Group leader
-
Marek Śmieja Unsupervised learning, tabular data, explainable AI
Post-docs
-
Kamil Książek Bioinformatics, explainable AI
Doctoral students
-
Marcin Przewięźlikowski Self-supervised learning, few-shot learning -
Andrzej Bedychaj Visual grounding, explainable AI -
Piotr Gaiński Cheminformatics, generative models -
Witold Wydmański Tabular data, large language models -
Patryk Marszałek Tabular foundation models, explainable AI -
Jan Masłowski Counterfactual explanations, computer vision -
Kacper Jurek Vision-language models, explainable AI
Former members
-
Bartosz Wójcik Efficient ML
-
Wojciech Batko Tabular data
-
Ewelina Jamrozik Cheminformatics
-
Magdalena Proszewska Generative models
-
Michał Znaleźniak Self-supervised learning
Selected publications
- Loading publications…
Recent grants
- Deep learning for tabular data
- Improving the transferability of self-supervised learning models
- Prototype networks as a step towards interpretable analysis of protein function
- Deep conditional generative models
- Deep processing of structured data
Join us
We take doctoral students through the Jagiellonian University doctoral school, and we supervise master's and bachelor's theses in the group's areas. If you want to work on any of the topics above, write with a short description of what interests you and what you have built so far.
Faculty of Mathematics and Computer Science
Jagiellonian University
6th Łojasiewicza Street
30-348 Kraków, Poland
Jagiellonian University
6th Łojasiewicza Street
30-348 Kraków, Poland