Projects
ml3

Selected Projects

Here you can find a small selection of our projects. We focus on spatio-temporal problems, such as user navigation on the web, adaptive testing and adaptive learning environments, and the coordination of football players on the pitch. If you are interested in getting to know more about or if you have any questions please do not hesitate to contact us.

Adaptive Learning in Economic Education: In the ALEE project, we explore adaptive (human) learning. We develop ML algorithms that are applied in a data sparse online learning setting and evaluate them in a field study in the area of economics education in German schools. This project is funded by the Joachim Herz Stiftung.

Masked Autoencoder for Multiagent Trajectories: We proposed the trajectory masked autoencoder (T-MAE), a self-supervised pretraining approach for multiagent trajectories. A workshop paper introducing the T-MAE can be accessed at https://link.springer.com. A full paper is currently under review. We will provide a PyTorch implementation at this location upon acceptance.

KISS_KI Simple and Scalable: KISS_KI aims at developing ML-based solutions for anomaly detection in the context of critical infrastructures. Different approaches based on robust Autoencoders and non-euclidean geometry are researched. It is funded by the Federal Ministry of Education and Research.

Animation of (Mouse) Trajectories: We created a tool to visualize spatio-temporal data like mouse movement data. You can toggle certain actions like clicks, keyboard hits or display the background. The switching between user and single sessions enhances the comparison.

[read more]

Synthi-Click: Anonymisation and Synthesis of Click-traces and Behaviour in the Web: The project aims to create synthetical dataset, matching the statistical properties of real movement data on German web pages using generative machine learning techniques. It is funded by the Federal Ministry of Education and Research.