Maximilian Herde

We propose a very performant graph-based neural operator architecture for learning the solution operator of PDEs from data on arbitrary domain discretizations, which has been tested on a challenging suite of benchmark datasets.