"Safe Active Learning" lecture with Christoph Zimmer from the Bosch Center for AI

After the opening event "How to attack Neural Networks" last year by Prof. Dr. Barbara Hammer, the members of the SAIL project welcomed their first external guest for the lecture series "Robust AI" on January 11, 2024.

Christoph Zimmer during his presentation "Smart Test Bench Automation with Safe Active Learning".

Dr. Christoph Zimmer, Lead Research Scientist at the Bosch Center for Artificial Intelligence, gave a one-hour lecture that could be followed both on site at the Technische Hochschule Ostwestfalen-Lippe (TH OWL) and online. The lecture was organized by the Research Group for Mathematics and Data Science at the Institute for Industrial Information Technology (inIT) under the direction of Prof. Markus Lange-Hegermann.

The title of the presentation was "Smart Test Bench Automation with Safe Active Learning". Model-based calculation plays a central role in many scientific fields. In recent years, data-based modeling has become increasingly important. In order to provide accurate data-based models, a sufficient amount of data must be collected. Depending on the technical system, this can be time-consuming and also safety-critical. Safe Active Learning is a sequential experimental design technique that automates data collection in an optimal and safe way. In his presentation, Dr. Christoph Zimmer introduced this method and showed how it can be used to operate a test bench in an industrial environment. The safe operating range is automatically learned and unwanted shutdowns of the test bench are avoided. This enables round-the-clock operation on weekdays and at weekends.

Around 100 participants, including members of the SAIL project as well as students and employees of TH OWL and the Institute for Industrial Information Technology (inIT), took part in the presentation and expanded their knowledge. The opportunity to follow the lecture online made it easier for many interested parties to participate and contributed to the success of the event.