Scaling GP-based PDE Solvers

MSc Research Project at the Methods of Machine Learning Group, supervised by Marvin Pförtner.

The project explored techniques to scale probabilistic PDE solvers to larger, more complex problems. The animation above demonstrates the result of this work: A simulation of the 1D shallow-water equations. This was not tractable within the previous framework.

I eventually continued this line of work into my PhD, which resulted in the GP-FVM paper.