PhD Data-driven learning of interpretable nonlinear models for high-tech systems

High-tech systems are becoming ever more complex due to increased requirements. Design and control of theses systems require accurate mathematical models that describe their complex nonlinear dynamical behavior. Engineers designing, building, and controlling these systems often have prior system knowledge. They desire to impose properties, e.g., stability, on the mathematical models and require these models to be interpretable.

This open PhD position aims to develop a framework that enables the identification of nonlinear mathematical models with pre-specified properties that are physically interpretable. To achieve this, the PhD candidate will exploit recent developments from the fields of machine learning, tensor algebra, system identification, and system theory to come to interpretable and stable models of complex dynamical systems.

Tasks

• Study the literature of nonlinear system identification, machine learning, tensor algebra, and system theory.
• Development of interpretable, data-driven nonlinear modeling approaches using tensor decompositions, nonlinear state-space identification techniques, and constrained Gaussian Processes.
• Analysis of the stability of the identified models in a time-efficient way.
• Enforcement of the stability during the identification step.
• Dissemination of the results of your research in international and peer-reviewed journals and conferences.
• Writing a successful dissertation based on the developed research and defending it.
• Assume educational tasks like the supervision of Master students during courses, internships, and graduation projects.