My research targets deep learning methods for physical simulations. The key question here is: How can we leverage our existing knowledge about numerical methods to improve our understanding of physical systems with the help of deep learning. In this context, I believe that neural networks are an exciting area for research, even more so when going beyond the established paths of pattern recognition problems. E.g., trained deep nets are able to learn structures of numerical errors of PDEs for which we have no analytical formulations, and they’re able to anticipate the dynamics of complex physical systems. There are many fascinating topics left to explore here.
If you’re interested in bachelor or master theses along these lines, please contact me. There are currently several interesting topics available in this area.
How to get in Touch
- eMail: nils.thuerey (at) tum.de
- Phone: +49 89 289 19484
- Fax: +49 89 289 19462
- Room: 02.13.061
A central focus of our work is coupling numerical simulations (e.g., Navier-Stokes solvers) with deep learning algorithms. An overview talk outlining some of our differentiable physics approaches can be found here:
Our Lagrangian approach for learning in unstructured mesh settings can be found here:
We’re additionally working on spatio-temporal (i.e. 4D) generative adversarial networks (GANs) for fluid simulations:
Our SIGGRAPH paper on flow descriptors from convolutional neural networks (numerical-viscosity-aware) for fluid flow:
For fun, here’s also an older animation which I created with our fluid control algorithm and Blender: