Fully Funded PhD: Dynamical Systems & the Dynamics of Learning

Why do high-dimensional neural networks often learn surprisingly simple representations? This project investigates the mathematical foundations of machine learning through nonlinear dynamics, invariant manifolds and spectral methods. Research topics include the geometry of neural-network training, implicit regularization, low-rank structures, model reduction, and the design of more efficient and interpretable learning algorithms.

What we offer

• Fully funded PhD position with competitive compensation
• Access to state-of-the-art high-performance computing and GPU infrastructure
• A research-intensive environment with minimal teaching responsibilities
• Opportunities for international collaborations, conferences and research visits
• A modern, international campus on the Red Sea

Who should apply

We welcome highly motivated candidates with strong quantitative backgrounds in mathematics, physics, computer science, engineering, statistics, scientific computing or related disciplines. Candidates should show mathematical curiosity, analytical ability and an interest in combining theoretical research with computational methods. Strong programming skills are particularly valuable. Prior expertise in every topic is not required.

How to apply

Interested candidates should send a CV, a short statement of research interests indicating this project, academic transcripts, and details of potential academic referees.

Start date: Fall 2027

Contact: Florian Kogelbauer, kopablauer@icloud.com