Follow-up on the work of Johannes Brandstetter*, Daniel Worrall*, Max Welling: Message Passing Neural PDE Solvers. This project builds upon their original repository: Link to code.
This repository contains experiments that evaluate the effect of incorporating a Gaussian smoothing kernel for weighting the messages from the neighbor nodes according to their proximity. The main changes are in experiments/models_gnn.py.