Below you can find a list of potential thesis topics (for most of the topics you should have taken my classes on scientific computing and / or GPU programming). I am also always willing to supervise related topics that fall into my expertise. Please get in touch with me if you are interested in a thesis.
- Neural network based matrix inversion: There are some theoretical results that describe the size of a neural network required to invert a d x d matrix. The topic of thesis would be to practically train a network on the task and compare the obtained results with the theoretical bounds.
- Local light field reconstruction with half precision (B.Sc. / M.Sc.)
The thesis develops an extension of local Fourier slice photography. There are various possible directions:
Extension of the algorithm so that it works (efficiently) with half precision data types. For a M.Sc. thesis this could be extended to a GPU implementation.
- Extension of the refocusing algorithm so that all-focus images can be generated.
- Extension of the algorithm to synthesise novel viewpoints directly from a compressed representation of the light field.
- Local all-focus light field reconstruction (M.Sc.)
The thesis develops an extension of local Fourier slice photography that develops an all-focus light field reconstruction technique based on the theory.
- GPU parallelization of Gascoigne (B.Sc. / M.Sc.)
The thesis explores efficient GPU parallelizations for the finite element fluid dynamics code Gascoigne 3d (in collaboration with Prof. Thomas Richter)
- Deep learning-based sub-grid scale models for (geophysical) fluid dynamics (B.Sc. / M.Sc.) The thesis explores the use of recurrent neural networks as subgrid scale models for (geophysical) fluid dynamics.
- Development of data base and web interface for fluid mechanics measurements (B.Sc.)This thesis is concerned with the structured organization and publication of a data set obtained at the Chair of Fluid dynamics.