Ekkehard Schnoor
Welcome to my academic website! I’m a PostDoc in the Machine Learning Research Group at Aalto University, led by Alex Jung, with a broad interest in the foundations of machine learning. Feel free to reach out if you are interested in my work!
📖 Bio
Before joining Aalto University, I was a member of the Explainable AI group at Fraunhofer HHI.
Prior to this, I obtained my PhD in mathematics from RWTH Aachen University under supervision of Holger Rauhut for my dissertation in statistical learning theory, at the intersection of compressive sensing and deep learning.
The overall goal of my research is to establish rigorous performance guarantees for high-dimensional machine learning, typically drawing on tools from high-dimensional probability and
leveraging on the concentration of measure phenomenon.
I'm also part of the team behind the Aalto Dictionary of Machine Learning.
My Erdös number is 4, via Holger Rauhut, Jürgen Prestin, and Charles Kam-tai Chui.
🔬 Research
I’m interested in a complementary view on machine learning, combining the perspectives of generalization and explainability, in a variety of interesting scenarios.
- Generalization: Including both non-asymptotic (e.g., using the Rademacher complexity) and asymptotic (based on random matrix theory) approaches.
- Explainability: In particular, Concept Activation Vectors (CAVs), and Sparse Autoencoders (SAEs) for Large Language Models (LLMs).
- Settings: Spanning a spectrum from basic linear models to deep neural networks, including networked models as in federated learning, and reinforcement learning.
- Techniques: Employing powerful tools from high-dimensional probability, such as chaining, random matrix theory (RMT), and (probabilistic) fixed-point methods.
📚 Selected Publications
For a complete list of publications, see my Google Scholar profile.
🏢 Contact
📍 Aalto University, Finland
📧 ekkehard.schnoor@aalto.fi
🔗 Google Scholar | GitHub | LinkedIn