A note on the physical interpretation of neural PDE's