Could you train a ML model that fills in the gaps in the fluid simulation like they do in ray tracing?