Karlson Pfannschmidt, M.Sc.

Research Focus

English: My research focus is on how to learn from context-dependent preferences, specifically choices. I investigate suitable representations for these very general preferences and evaluate their quality empirically. This research has applications in information retrieval (such as which search results a user clicks on) and eCommerce (such as which products a user finds interesting and ultimately buys). Currently, I am working in the EFRE-project "Smart-GM", where we are developing software that uses crowdsourcing and preference learning to assist users in business model innovation.

German: Mein Forschungsschwerpunkt liegt darauf, aus kontextabhängigen Präferenzen, insbesondere Entscheidungen, zu lernen. Ich untersuche geeignete Darstellungen für diese sehr allgemeinen Präferenzen und evaluieren ihre Qualität empirisch. Diese Forschung hat Anwendungen in der Informationssuche (wie welche Suchergebnisse ein Benutzer anklickt) und im elektronischen Handel (wie welche Produkte ein Benutzer interessant findet und schließlich kauft). Derzeit arbeite ich im EFRE-Projekt "Smart-GM", wo wir eine Software entwickeln, die Nutzer bei der Geschäftsmodell-Innovation mittels Crowdsourcing und Präferenzlernen unterstützt.

Research Area

Preference Learning

Conference Talks and Posters

• A Characterization of Choice Functions Representable by Pareto-Embedding of Alternatives, DA2PL 2022, Compiègne, France, 2022 (talk + extended abstract)
• Learning Choice Functions via Pareto-Embeddings, KI 2020, Bamberg, Germany, 2020 (talk + Paper)
• Shapley Curves: A New Concept for Modelling Feature Importance, LWDA 2017, Rostock, Germany, 2017 (Poster + Abstract)
• Evaluating Tests in Medical Diagnosis: Combining Machine Learning with Game-Theoretical Concepts, IPMU 2016, Eindhoven, The Netherlands, 2016 (talk + Paper)

Publications

• Karlson Pfannenschmidt, Pritha Gupta, Björn Haddenhorst, Eyke Hüllermeier (2022)
  Learning context-dependent choice functions
  In: International Journal of Approximate Reasoning, Vol. 140, January 2022: pp. 116-155
• Michael Dellnitz, Eyke Hüllermeier, Marvin Lücke, Sina Ober-Blöbaum, Christian Offen, Sebastian Peitz, Karlson Pfannenschmidt (2021)
  Efficient time stepping for numerical integration using reinforcement learning
  In: ArXiv abs/2104.03562: n. pag.
• Karlson Pfannenschmidt, Pritha Gupta, Eyke Hüllermeier, (2018)
  Deep Architectures for Learning Context-dependent Ranking Functions
  In: ArXiv abs/1803.05796: n. pag.
• Tim Head, MechCoder, Gilles Louppe, Iaroslav Shcherbatyi, fcharras, Zé Vinícius, cmmalone, Christopher Schröder, nel215, Nuno Campos, Todd Young, Stefano Cereda, Thomas Fan, Justus Schwabedal, Hvass-Labs, Mikhail Pak, SoManyUsernamesTaken, Fred Callaway, Loïc Estève, Lilian Besson, Peter M. Landwehr, Pavel Komarov, Mehdi Cherti, Kejia (KJ) Shi, Karlson Pfannenschmidt, Fabian Linzberger, Christophe Cauet, Anna Gut, Andreas Mueller, Alexander Fabisch, (2018)
  scikit-optimize/scikit-optimize: High five - v0.5
  
• Pritha Gupta, Alexander Hetzer, Tanja Tornede, Sebastian Gottschalk, Andreas Kornelsen, Sebastian Osterbrink, Karlson Pfannenschmidt, Eyke Hüllermeier, (2017)
  jPL: A java-based software framework for preference learning
  In: Proceedings of the LWDA
• Kalina Jasinska, Krzysztof Dembczynski, Róbert Busa-Fekete, Karlson Pfannenschmidt, Timo Klerx, Eyke Hüllermeier, (2016)
  Extreme F-Measure Maximization using Sparse Probability Estimates
  In: International Conference on Machine Learning 33, pp. 1435-1444