ECML PKDD 2026 as a Platform for Lamarr Research and Networking

Group photo of Lamarr network members wearing conference badges in front of the ECML PKDD 2026 sponsor wall in Naples.

The Lamarr Institute participated in ECML PKDD 2026 in Naples with research contributions on explainable, resource-aware and robust Artificial Intelligence and as a Gold Sponsor. From September 7 to 11, researchers from the Institute presented their work at the European conference on Machine Learning and Knowledge Discovery. The Institute’s joint presence brought the contributions together under one visible profile and created additional opportunities for exchange and networking. The work addressed fundamental questions about how learning systems process information, how their decisions can be understood and how reliably they perform under changing conditions.

A Broad Spectrum of Current ML Research

The Lamarr contributions combined fundamental methodological research with applications in medicine and data analysis. Lorenz Sparrenberg investigated geometric changes in the internal representations of transformer models. In several contributions from the field of Visual Analytics, Natalia Andrienko and Gennady Andrienko demonstrated how the logic of rule-based models and interactions in time series can be examined. Further work addressed continuous graph learning, the detection of epileptic seizures and robust analyses of wearable data. Kirsten Köbschall and Sebastian Buschjäger presented resource-aware Online Learning on heterogeneous devices, while Augustina Chidinma Amakor addressed the interactive balancing of competing objectives in multi-task models.

Reliable Models Under Real-World Conditions

The different studies address shared challenges in current AI research: How do models develop meaningful representations? How can their decisions be understood? And how can they remain reliable when data, signals or application contexts change? Lamarr Co-Director Prof. Dr. Stefan Wrobel discussed these questions in his conference presentation and placed them within the Institute’s broader research context.

Changes in data distributions were also the focus of the workshop “Quantification and Classification under Dataset Shift”, co-organized by Lamarr research associate Mirko Bunse. Dataset shift occurs when the data encountered during deployment differ from the data used to train a model. This can happen, for example, when a medical system is applied to new patient groups or when sensors produce different signals under changing conditions. The workshop brought together researchers and practitioners to discuss robust methods for such situations and their practical applications.

The Lamarr Institute at ECML PKDD 2026: research presentations, networking and exchange at the Institute’s booth in Naples.

Scientific Visibility and On-Site Networking

The joint Lamarr booth made the research contributions visible beyond the individual sessions and offered international early-career researchers insights into the Institute’s research profile and potential exchange formats. In addition, Lamarr created dedicated networking opportunities outside the official conference program. Partners and researchers from different locations and research areas strengthened existing contacts, established new connections and identified opportunities for future collaboration. In this way, the Institute’s presence in Naples combined international visibility with scientific networking within and beyond the Lamarr network.

More news