Research

We are drawn to deep priors and structural inductive biases—the kind that bridge disciplines, inspire new mathematics, and open novel research fields. Many of our projects sit at the cusp of theory and application, where tools from geometry, topology or quantum physics can enable new forms of perception, generation, or understanding. As a result, we publish in a variety of areas that cross-feed each other, including: 3D Computer Vision including 3D/4D Generative Priors, Geometric & Topological Deep Learning, Statistical & Topological Learning Theory, Mechanistic Interpretability, Quantum Computer Vision and Generative Models for Biochemistry.

Core Areas

Topological Deep Learning A new way to work with higher-order, complex data in the context of deep learning.

3D/4D Computer Vision Our generative models allow for reconstruction, analysis and sampling of 3D/4D shapes.

Learning Theory A new suite of statistical learning theory informed by the topology of representations.

Exploratory Research Across Disciplines

AI-driven
Structural Biology Topological deep learning enables higher-order modeling of molecules and generative AI in structural biology.

Quantum Computer Vision & Machine Learning A new computational paradigm for addressing vision and learning problems.

Next: Cognitive Sciences and Brain Modelling Our next research agenda involves uncovering the higher-order nature of cognition via TDL on brain dynamics.