Prof. Dr. Markus Lange-Hegermann, member of the inIT Executive Board, co-organised a workshop on ‘Learning from Small Data’ at ECML PKDD 2025 and presented two scientific papers. The activities at the conference were part of the SAIL project.
ECML PKDD brings over 1,300 experts to Porto
The European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD) is the most important European conference on machine learning and data mining. In addition to keynotes by leading scientists, the programme includes specialist lectures, poster sessions, a Discovery Challenge, industry and doctoral tracks, and numerous workshops. More than 1,300 experts from academia and industry gathered in Porto to discuss current research.
Workshop: Strategies for data-poor scenarios
One item on the conference agenda was the Learning from Small Data (LFSD 2025) workshop, co-organised by Prof. Dr. Markus Lange-Hegermann. Researchers from various disciplines discussed strategies for developing reliable models even with small amounts of data – a topic of high relevance for industrial applications. Among other things, approaches such as few-shot learning, data augmentation, and physics-informed machine learning were discussed, with a particular focus on practical methods for real-world applications.
Scientific contributions from inIT
inIT was represented in the workshop with two technical papers:
Together with his research associate Andreas Besginow, Prof. Dr. Markus Lange-Hegermann presented the paper ‘Varying Informativeness of Inductive Bias in Gaussian Processes Regression for Small Data’. It examines how different induction assumptions influence the performance of Gaussian process models in data-poor scenarios.
He also presented the paper ‘Physics-Informed Diffusion Models for Unsupervised Anomaly Detection in Multivariate Time Series’ together with Juhi Soni from Fraunhofer IOSB-INA. This work shows how physical knowledge can be used in diffusion models to reliably detect anomalies in complex time series.
In addition to Prof. Dr. Markus Lange-Hegermann and Andreas Besginow, SAIL colleagues Prof. Dr. Wolfram Schenck and Dr. Alaa Tharwat from Bielefeld University of Applied Sciences also took part in the conference.
Research results in international dialogue
With its involvement in the ECML PKDD, inIT is highlighting the international visibility of its research into data-driven methods. Prof. Dr. Markus Lange-Hegermann emphasises:
‘The challenges of learning with small amounts of data are highly relevant in many areas of industry. The workshop showed how new approaches and international cooperation can lead to sustainable solutions. The exchange with experts from all over the world who came to the conference in Porto was particularly enriching.’