Theses
The following thesis topics are currently available. Depending on the degree (BSc. / MSc.), the scope may vary in terms of depth and complexity
The task of this thesis is to give a detailed introduction to the concept of geometric deep learning (for more details, see this course on geometric deep learning or this book on equivariant convolutional networks). A particular focus should be put on the practical implementation of steerable networks using the ESCNN library, with applications to physics simulations data.
An introduction to Kolmogorov-Arnold Networks (KANs) and application to the prediction of dynamical systems.
Extension of a symmetric reinforcemet learning framework (known as homomorphic MDPs) from discrete symmetries (flips, 90° rotations) to continuous ones, with applications in robotics or continuum/fluid mechanics.