| Research Interests | Explainable Artificial Intelligence (XAI) Hierarchical analysis of factor graphs Statistical boosting algorithms Methods for robust regression |
| Current Talks | • Compression vs. Accuracy: Compact Models for Efficiency and Interpretation. Tutorial at the IJCAI-ECAI-26 35th International Joint Conference on Artificial Intelligence - 29th European Conference on Artificial Intelligence, Bremen Slides Link to event • Inducing Comparability of Factorised Probability Distributions. 3rd Logic and AI Meet-up, Hagen • Compression versus Accuracy: A Hierarchy of Lifted Models. 28th European Conference on Artificial Intelligence (ECAI 2025), Bologna Slides • Towards Explainability of Approximate Lifted Model Construction: A Geometric Perspective. Joint Workshop on Humanities-Centred Artificial Intelligence and Formal & Cognitive Reasoning co-located with 48th German Conference on Artificial Intelligence, Potsdam Link to event |
| Current Publications | • Braun, Tanya; Gehrke, Marcel; Sauerwald, Kai; Speller, Jan Variable Elimination for Forgetting in Probabilistic Graphical Models - A Work in Progress. Joint proceedings of the Workshop on Theory and Methods for Abstraction (THEMA 2026) and the Workshop on Modularity and Splitting Techniques for Knowledge Representation and Reasoning (MoST 2026) at the 9th Federated Logic Conference (FLoC 2026), 2026 online • Luttermann, Malte; Speller, Jan; Gehrke, Marcel; Braun, Tanya; Möller, Ralf; Hartwig, Mattis Approximate Lifted Model Construction. IJCAI-25 Proceedings of the 34th International Joint Conference on Artificial Intelligence, 2025 online • Speller, Jan; Luttermann, Malte; Gehrke, Marcel; Braun, Tanya Compression versus Accuracy: A Hierarchy of Lifted Models. ECAI-25 Proceedings of the 28th European Conference on Artificial Intelligence Vol. Volume 413, 2025 online • Speller, Jan; Luttermann, Malte; Gehrke, Marcel; Braun, Tanya Towards Explainability of Approximate Lifted Model Construction: A Geometric Perspective. Proceedings of the Joint Workshop on Humanities-Centred Artificial Intelligence and Formal & Cognitive Reasoning co-located with 48th German Conference on Artificial Intelligence, 2025 online • Ehrentraut, H; Massoth, G; Delis, A; Thewes, B; Hoch, J; Majchrzak, M; Weber-Schehl, M; Mayr, A; Abulizi I; Speller, J; Meybohm, P; Steinisch, A; Koessler, J; Strauss, AC; Wittmann, M; Velten, M Implications of packed red bloods cells production and transfer on post transfusion hemoglobin increase. Journal of Clinical Anesthesia Vol. 102, 2025, pp 111743 online • Angiuli, V; Rohner, M; Wittmann, M; Piekarski, F; Kruse, J; Silaschi, M; Kunsorg, A; Mayr, A; Abulizi, I; Speller, J; Bakhtiary, F; Velten, M Ultra-fast extubation following cardiac surgery improves hemodynamic stability and reduces ICU workload. Frontiers in Cardiovascular Medicine Vol. 12, 2025, pp 1695955 online • Siener, R; Ernsten, C; Speller, J; Scheurlen, C; Sauerbruch, T; Hesse, A Intestinal Oxalate Absorption, Enteric Hyperoxaluria, and Risk of Urinary Stone Formation in Patients with Crohn's Disease. Nutrients Vol. 16 (2), 2024 online • Speller, Jan; Staerk, Christian; Gude, Francisco; Mayr, Andreas; Robust gradient boosting for generalized additive models for location, scale and shape. Advances in Data Analysis and Classification, 2023 online • Mayr, Andreas; Wistuba, Tobias; Speller, Jan; Gude, Francisco; Hofner, Benjamin; Linear or smooth? Enhanced model choice in boosting via deselection of base-learners. Statistical Modelling Vol. 23 (5-6), 2023 online |
| Current Projects | • HAPPI - Human-aware PGMs and Probabilistic Inference via Lifted Model Reconciliation Bei diesem Projekt geht es darum, die Kommunikation zwischen KI-Systemen und Menschen zu verbessern. In KI-basierten Systemen werden häufig Modelle aus Daten gelernt. Diese Modelle können zwar auch Wissen von Expertinnen und Experten beinhalten, sind aber häufig durch die gelernten Informationen aus den Daten nicht mehr einfach zu erklären. In dem Projekt wird dort angesetzt, indem Methoden erforscht werden, wie solche gelernten Modelle mit den Erwartungen von Expertinnen und Experten in Übereinstimmung gebracht werden können und Mensch und System sich gegenseitig ihre Wissenslücken erklären können. Auf lange Sicht sollen diese Systeme zum Beispiel in Medizin oder Geisteswissenschaften ihre Vorschläge erklären können. | jan dot speller at uni-muenster dot de |
| Phone | +49 251 83-33747 |
| Room | 603 |
| Secretary | Sekretariat Steinhoff Frau Gerlinde Steinhoff Telefon +49 251 83-38447 Zimmer 602 |
| Address | Herr Dr. Jan Speller Institut für Informatik Fachbereich Mathematik und Informatik der Universität Münster Einsteinstrasse 62 48149 Münster Deutschland |
| Diese Seite editieren |
