Machine Learning group
The (probabilistic) machine learning group is led by Isabel Valera, Professor of Machine Learning at Saarland University, Adjunct Faculty of the MPI-SWS and research fellow of the European Laboratory for Learning and Intelligent Systems (ELLIS).
We develop cutting-edge trustworthy machine learning methods to be deployed in the real-world. Our research can be broadly categorized in three main topics: fair, interpretable and robust machine learning. We are an active and diverse research team, with interests in a wide range of ML approaches including deep learning, probabilistic modeling, causal inference, time series analysis, and many more.
Our research has a strong societal component and can be applied in a broad range of application domains, from medicine and psychiatry to social and communication systems. As an example, our recent research has focused on algorithmic decision making in several domains, including hiring processes, pre-trial bail, or loan approval.
Auf Deutsch:
Die Machine Learning Gruppe wird von Isabel Valera geführt, die sowohl Professorin für Machine Learning and der Universität des Saarlandes (UdS), auch zur Fakultät des MPI-SWS gehört, und Research Fellow des European Laboratory for Learning and Intelligent Systems (ELLIS) ist.
Die Gruppe entwickelt innovative und vertrauenswürdige Machine Learning Methoden für den Einsatz in der echten Welt. Unsere Forschung kann grob in drei Hauptbereiche eingeteilt werden: faires, interpretierbares, bzw. robustes maschinelles lernen. Wir sind ein aktive und diverses Forschungsteam, und unsere Interessen umfassen eine große Auswahl von Herangehensweisen an das Maschinelle Lernen, wie etwa Deep Learning, probabilistische Modellierungen, Causal Inference, time series analysis, und weitere.
Unsere Forschung hat eine starke gesellschaftliche Komponente und kann in einer großen Auswahl an Anwendungsgebieten eingesetzt werden. Beispielsweise befasst sich unsere Forschung mit dem algorithmischen Treffen von Entscheidungen das in Gebieten wie dem Einstellungsprozess oder Darlehnsantragsbewilligungen stattfindet.
News
Paper “Weakly Supervised Detection of Hallucinations in LLM Activations” accepted to NeurIPS Workshop on Socially Responsible Language Modelling Research
The paper "Weakly Supervised Detection of Hallucinations in LLM Activations" coauthored by our PhD Student Miriam Rateike was accepted to the NeurIPS23 Workshop "Socially Responsible Language Modelling Research" (https://solar-neurips.github.io/). The full list of...
Abstract by Miriam Rateike accepted to WiML Workshop at NeurIPS23
The abstract "Exploring Unregulated Discrimination: Algorithmically Created Disadvantaged Groups" by our PhD Student Miriam Rateike was accepted to the WiML workshop at NeurIPS23. More information on the workshop can be found on this and this website.
PhD student Miriam Rateike received a Google PhD Fellowship award
Google has awarded a prestigious Google PhD Fellowship to our PhD student Miriam Rateike. The Google PhD Fellowship program recognizes and supports exceptional graduate students in computer science and related disciplines. More details can be found via her Twitter...