PDE-G-CNNs utilize solvers of geometrically meaningful evolution PDEs for improved performance in neural networks.
