News
Robust data analysis and reliable detection: Michael Muma presented his research
The International Workshop on Resilient 6G Networks took place in Bruges in early September as part of the European Signal Processing Conference. Michael Muma, a professor at the HDCC, gave a presentation. | © AdobeStockArcady
How can relevant information be reliably extracted from high-dimensional sensor data? Michael Muma, Professor at the Hessian Drone Competence Centre (HDCC) and Head of the Robust Data Science Group at TU Darmstadt, addresses questions such as this and recently presented his latest research on robust data-driven signal processing and machine learning at two specialist events in Bruges and Darmstadt.
„Robust data analysis, reliable detection and resilient communication are key building blocks that are also being used in our project at the HDCC,“ explains Michael Muma. At the International Workshop on Resilient 6G Networks (WresNet 6G), which took place on 4 September in Bruges, Belgium, as part of the European Signal Processing Conference (EUSIPCO), Professor Michael Muma delivered a presentation entitled „From Robust Signal Processing to Trustworthy and Scalable Learning for High-Dimensional Sensor Data“. The presentation focused in particular on the question of how data-driven methods can deliver reliable results even under difficult conditions.
Michael Muma gave a further talk on 25 September at the meeting of the ITG specialist group on „Signal Processing and Machine Learning“ at TU Darmstadt. Under the title „From Robust Signal Processing to Trustworthy Scalable Learning: Statistical Guarantees from Sensor Data to Genomics“, he presented the latest methodological developments at the interface between robust signal processing, statistics and machine learning.
Reliable variable selection with T-Rex
A key focus of both presentations was the T-Rex framework, which the researchers in Professor Muma’s research group have developed for reliable variable selection in high-dimensional data. The aim is to reliably identify, from a vast number of possible influencing factors, those that are actually relevant, whilst statistically controlling for false positives. „This is particularly important when data is noisy, incomplete or influenced by outliers,“ emphasises Michael Muma.
Significance for resilient 6G networks and the HDCC
At the International Workshop on Resilient 6G Networks, there was a particular focus on future 6G networks. Such networks will also be relevant for connected drone systems, drone detection and applications in the field of critical infrastructure. They must function reliably even in the face of disruptions, failures, attacks or rapidly changing environmental conditions. Robust statistical methods can help here by identifying relevant patterns in network and sensor data, detecting anomalies and making data-driven decisions more transparent and resilient.
At the ITG specialist group meeting, however, the focus was on methodological aspects – namely, the integration of traditional robust signal processing, modern statistics and machine learning. Professor Muma reported on the latest enhancements to the T-Rex framework, including those for complex data structures, networked systems and robust decision-making processes.
The research and methods presented here are also of potential interest in the context of HDCC issues. In particular, when analysing sensor, communication and situational awareness data, they can help to obtain reliable information and develop resilient systems.