Sophie Neubauer & Philipp Neubauer | DatenVorsprung GmbH
Lecture language: English
Neural networks (NNs) are increasingly deployed to control complex systems where traditional control approaches are not capable to do so. When controlling complex systems, for example embedded systems in robotics, the big open question is: How to provide safety-guarantees, predictable behavior, strong assurances, and thus trustworthiness, so that the use of NN controllers becomes a feasible strategy? The main focus of this talk to show you how to build NN-controlled systems that consistently perform as intended, with predictable behavior and robustness against various conditions or inputs. This includes systematic testing and statistical reachability analysis to ensure reliability in real-world. In reachability analysis, this question is addressed by giving over-approximations of the solution space of the AI systems, considering the desired accuracy and computational power. The goal is to ensure that trustworthy AI becomes the cornerstone for empowering control over complex systems.
Sophie Neubauer, a mathematician, completed her doctoral program in the Cyber-Physical Systems Group at TU Wien and is currently CTO of DatenVorsprung. Her passion is to combine her mathematical skills with Artificial Intelligence with the goal to provide trustworthy AI in safety-critical environments.
Philipp Neubauer, the CEO of DatenVorsprung, previously held the position of Head of Automation at Cubicure GmbH. His leadership played a pivotal role in transforming a startup in additive manufacturing into a globally renowned company specializing in mechanical engineering and process technology. With his expertise in system architectures, Philipp is now focused on enhancing the scalability of the company’s systems.
Mehr Infos zur SAINT:
saint.fhstp.ac...
#fhstp #SAINT #ai
19 сен 2024
