Dr. Johannes Haug
After completing my Master's degree in Business Informatics at TUM (with a previous Bachelor's degree at DHBW Stuttgart), I started as a research assistant and PhD student in the Data Science and Analytics group at the University of Tübingen in October 2018. In September 2022, I finished my PhD project "Towards Reliable Machine Learning in Evolving Data Streams" under the supervision of Prof. Gjergji Kasneci. In October 2022, I joined the Bosch Center for Artificial Intelligence (BCAI).
Research Interests
- Feature Selection in Data Streams/ Online Feature Selection
- Stability/Robustness of Feature Selection Algorithms
- Concept Drift Detection and Adaptation
- Local Additive Attributions and Explanations in Dynamic Data Streams
- Efficient, Powerful and Interpretable Online Predictive Models
- Meaningful and Standardized Evaluation of Online Learning Methods
Detailed information on our research work is available here.
Publications
2022
Deep Neural Networks and Tabular Data: A Survey by Vadim Borisov, Tobias Leemann, Kathrin Sessler, Johannes Haug, Martin Pawelczyk and Gjergji Kasneci accepted to IEEE Transactions on Neural Networks and Learning Systems (TNNLS).
Towards Reliable Machine Learning in Evolving Data Streams by Johannes Haug was published as a dissertation by the Faculty of Science of the University of Tübingen.
Change Detection for Local Explainability in Evolving Data Streams by Johannes Haug, Alexander Braun, Stefan Zürn, and Gjergji Kasneci has been accepted at CIKM 2022 (github).
Standardized Evaluation of Machine Learning Methods for Evolving Data Streams by Johannes Haug, Effi Tramountani and Gjergji Kasneci was published on arXiv. The corresponding float-evaluation framework (Python) can be found on Github or Pypi.
Dynamic Model Tree for Interpretable Data Stream Learning by Johannes Haug, Klaus Broelemann and Gjergji Kasneci was accepted at ICDE 2022 (github).
2021
Deep Neural Networks and Tabular Data: A Survey by Vadim Borisov, Tobias Leemann, Kathrin Seßler, Johannes Haug, Martin Pawelczyk and Gjergji Kasneci was published on arXiv.
TüEyeQ, a rich IQ test performance data set with eye movement, educational and socio-demographic information by Enkelejda Kasneci, Gjergji Kasneci, Tobias Appel, Johannes Haug, Franz Wortha, Maike Tibus, Ulrich Trautwein and Peter Gerjets was accepted at the Nature Scientific Data Journal (8, 2021) (dataset, github).
On Baselines for Local Feature Attributions by Johannes Haug, Stefan Zürn, Peter El-Jiz and Gjergji Kasneci was accepted at the Workshop on Explainably Agency in AI @AAAI 2021 (Abstract, .bib-file, github).
2020
Learning Parameter Distributions to Detect Concept Drift in Data Streams by Johannes Haug and Gjergji Kasneci was accepted at ICPR 2020 (Abstract, .bib-file, github).
Leveraging Model Inherent Variable Importance for Stable Online Feature Selection by Johannes Haug, Martin Pawelczyk, Klaus Broelemann and Gjergji Kasneci was accepted at KDD 2020 (Abstract, .bib-file, github).
2019
Towards User Empowerment by Martin Pawelczyk, Johannes Haug, Klaus Broelemann and Gjergji Kasneci was accepted at the HCML Workshop @NeurIPS 2019 (Abstract, bib-file).
CancelOut: A Layer for Feature Selection in Deep Neural Networks by Vadim Borisov, Johannes Haug und Gjergji Kasneci was accepted at ICANN 2019 (Abstract, .bib-file).