PhD Position in Formal Methods for Learned Systems

Updated: over 2 years ago
Deadline: 18 Sep 2021

The main difference between most modern learned systems and traditional automation is that behavior is not programmed, but derived from examples. When using such systems in practice we need to be able to identify when the system may not be doing what we want. The challenge this project addresses is to support experts in monitoring learned systems, so that they can step in to prevent AI systems making decisions that are discriminating, dangerous or simply wrong.

We consider human feedback to be essential for achieving reliability of learned systems in the real world. We believe in the hidden potential of the synergies between formal and learning-based methods. You will conduct both theoretical and empirical research on monitoring learned systems. You can choose to primarily focus on one of the aspects or their intersection. We envision the project to have both significant scientific and practical impact, across areas that use learning for automation.

You will be part of the Algorithmics Group in the Department of Software Technology of the Faculty of Electrical Engineering, Mathematics and Computer Science. In the Algorithmics group, we aim to design, and understand fundamental properties of, planning and coordination algorithms for intelligent decision making in real world applications. As the member of the group, you will reinforce and extend the group's research activities in the emergent interdisciplinary field combining formal methods and machine learning. Algorithmics group has a long-standing track record in the area of artificial intelligence. Our research output is supported by several awards and by numerous conference papers at IJCAI, AAAI, AAMAS, ECAI, as well as TACAS, CPSWeek, ATVA. In addition, you will have an opportunity to collaborate with the Interactive Intelligence Group internationally recognized for their research on socially interactive agents



Similar Positions